net/ipv4/ipmr.c

   1 /*
   2  *      IP multicast routing support for mrouted 3.6/3.8
   3  *
   4  *              (c) 1995 Alan Cox, <alan@lxorguk.ukuu.org.uk>
   5  *        Linux Consultancy and Custom Driver Development
   6  *
   7  *      This program is free software; you can redistribute it and/or
   8  *      modify it under the terms of the GNU General Public License
   9  *      as published by the Free Software Foundation; either version
  10  *      2 of the License, or (at your option) any later version.
  11  *
  12  *      Fixes:
  13  *      Michael Chastain        :       Incorrect size of copying.
  14  *      Alan Cox                :       Added the cache manager code
  15  *      Alan Cox                :       Fixed the clone/copy bug and device race.
  16  *      Mike McLagan            :       Routing by source
  17  *      Malcolm Beattie         :       Buffer handling fixes.
  18  *      Alexey Kuznetsov        :       Double buffer free and other fixes.
  19  *      SVR Anand               :       Fixed several multicast bugs and problems.
  20  *      Alexey Kuznetsov        :       Status, optimisations and more.
  21  *      Brad Parker             :       Better behaviour on mrouted upcall
  22  *                                      overflow.
  23  *      Carlos Picoto           :       PIMv1 Support
  24  *      Pavlin Ivanov Radoslavov:       PIMv2 Registers must checksum only PIM header
  25  *                                      Relax this requirement to work with older peers.
  26  *
  27  */
  28
  29 #include <asm/uaccess.h>
  30 #include <linux/types.h>
  31 #include <linux/capability.h>
  32 #include <linux/errno.h>
  33 #include <linux/timer.h>
  34 #include <linux/mm.h>
  35 #include <linux/kernel.h>
  36 #include <linux/fcntl.h>
  37 #include <linux/stat.h>
  38 #include <linux/socket.h>
  39 #include <linux/in.h>
  40 #include <linux/inet.h>
  41 #include <linux/netdevice.h>
  42 #include <linux/inetdevice.h>
  43 #include <linux/igmp.h>
  44 #include <linux/proc_fs.h>
  45 #include <linux/seq_file.h>
  46 #include <linux/mroute.h>
  47 #include <linux/init.h>
  48 #include <linux/if_ether.h>
  49 #include <linux/slab.h>
  50 #include <net/net_namespace.h>
  51 #include <net/ip.h>
  52 #include <net/protocol.h>
  53 #include <linux/skbuff.h>
  54 #include <net/route.h>
  55 #include <net/sock.h>
  56 #include <net/icmp.h>
  57 #include <net/udp.h>
  58 #include <net/raw.h>
  59 #include <linux/notifier.h>
  60 #include <linux/if_arp.h>
  61 #include <linux/netfilter_ipv4.h>
  62 #include <linux/compat.h>
  63 #include <linux/export.h>
  64 #include <net/ip_tunnels.h>
  65 #include <net/checksum.h>
  66 #include <net/netlink.h>
  67 #include <net/fib_rules.h>
  68 #include <linux/netconf.h>
  69
  70 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
  71 #define CONFIG_IP_PIMSM 1
  72 #endif
  73
  74 struct mr_table {
  75         struct list_head        list;
  76 #ifdef CONFIG_NET_NS
  77         struct net              *net;
  78 #endif
  79         u32                     id;
  80         struct sock __rcu       *mroute_sk;
  81         struct timer_list       ipmr_expire_timer;
  82         struct list_head        mfc_unres_queue;
  83         struct list_head        mfc_cache_array[MFC_LINES];
  84         struct vif_device       vif_table[MAXVIFS];
  85         int                     maxvif;
  86         atomic_t                cache_resolve_queue_len;
  87         bool                    mroute_do_assert;
  88         bool                    mroute_do_pim;
  89 #if defined(CONFIG_IP_PIMSM_V1) || defined(CONFIG_IP_PIMSM_V2)
  90         int                     mroute_reg_vif_num;
  91 #endif
  92 };
  93
  94 struct ipmr_rule {
  95         struct fib_rule         common;
  96 };
  97
  98 struct ipmr_result {
  99         struct mr_table         *mrt;
 100 };
 101
 102 /* Big lock, protecting vif table, mrt cache and mroute socket state.
 103  * Note that the changes are semaphored via rtnl_lock.
 104  */
 105
 106 static DEFINE_RWLOCK(mrt_lock);
 107
 108 /*
 109  *      Multicast router control variables
 110  */
 111
 112 #define VIF_EXISTS(_mrt, _idx) ((_mrt)->vif_table[_idx].dev != NULL)
 113
 114 /* Special spinlock for queue of unresolved entries */
 115 static DEFINE_SPINLOCK(mfc_unres_lock);
 116
 117 /* We return to original Alan's scheme. Hash table of resolved
 118  * entries is changed only in process context and protected
 119  * with weak lock mrt_lock. Queue of unresolved entries is protected
 120  * with strong spinlock mfc_unres_lock.
 121  *
 122  * In this case data path is free of exclusive locks at all.
 123  */
 124
 125 static struct kmem_cache *mrt_cachep __read_mostly;
 126
 127 static struct mr_table *ipmr_new_table(struct net *net, u32 id);
 128 static void ipmr_free_table(struct mr_table *mrt);
 129
 130 static int ip_mr_forward(struct net *net, struct mr_table *mrt,
 131                          struct sk_buff *skb, struct mfc_cache *cache,
 132                          int local);
 133 static int ipmr_cache_report(struct mr_table *mrt,
 134                              struct sk_buff *pkt, vifi_t vifi, int assert);
 135 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
 136                               struct mfc_cache *c, struct rtmsg *rtm);
 137 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
 138                                  int cmd);
 139 static void mroute_clean_tables(struct mr_table *mrt);
 140 static void ipmr_expire_process(unsigned long arg);
 141
 142 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
 143 #define ipmr_for_each_table(mrt, net) \
 144         list_for_each_entry_rcu(mrt, &net->ipv4.mr_tables, list)
 145
 146 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
 147 {
 148         struct mr_table *mrt;
 149
 150         ipmr_for_each_table(mrt, net) {
 151                 if (mrt->id == id)
 152                         return mrt;
 153         }
 154         return NULL;
 155 }
 156
 157 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
 158                            struct mr_table **mrt)
 159 {
 160         struct ipmr_result res;
 161         struct fib_lookup_arg arg = { .result = &res, };
 162         int err;
 163
 164         err = fib_rules_lookup(net->ipv4.mr_rules_ops,
 165                                flowi4_to_flowi(flp4), 0, &arg);
 166         if (err < 0)
 167                 return err;
 168         *mrt = res.mrt;
 169         return 0;
 170 }
 171
 172 static int ipmr_rule_action(struct fib_rule *rule, struct flowi *flp,
 173                             int flags, struct fib_lookup_arg *arg)
 174 {
 175         struct ipmr_result *res = arg->result;
 176         struct mr_table *mrt;
 177
 178         switch (rule->action) {
 179         case FR_ACT_TO_TBL:
 180                 break;
 181         case FR_ACT_UNREACHABLE:
 182                 return -ENETUNREACH;
 183         case FR_ACT_PROHIBIT:
 184                 return -EACCES;
 185         case FR_ACT_BLACKHOLE:
 186         default:
 187                 return -EINVAL;
 188         }
 189
 190         mrt = ipmr_get_table(rule->fr_net, rule->table);
 191         if (mrt == NULL)
 192                 return -EAGAIN;
 193         res->mrt = mrt;
 194         return 0;
 195 }
 196
 197 static int ipmr_rule_match(struct fib_rule *rule, struct flowi *fl, int flags)
 198 {
 199         return 1;
 200 }
 201
 202 static const struct nla_policy ipmr_rule_policy[FRA_MAX + 1] = {
 203         FRA_GENERIC_POLICY,
 204 };
 205
 206 static int ipmr_rule_configure(struct fib_rule *rule, struct sk_buff *skb,
 207                                struct fib_rule_hdr *frh, struct nlattr **tb)
 208 {
 209         return 0;
 210 }
 211
 212 static int ipmr_rule_compare(struct fib_rule *rule, struct fib_rule_hdr *frh,
 213                              struct nlattr **tb)
 214 {
 215         return 1;
 216 }
 217
 218 static int ipmr_rule_fill(struct fib_rule *rule, struct sk_buff *skb,
 219                           struct fib_rule_hdr *frh)
 220 {
 221         frh->dst_len = 0;
 222         frh->src_len = 0;
 223         frh->tos     = 0;
 224         return 0;
 225 }
 226
 227 static const struct fib_rules_ops __net_initconst ipmr_rules_ops_template = {
 228         .family         = RTNL_FAMILY_IPMR,
 229         .rule_size      = sizeof(struct ipmr_rule),
 230         .addr_size      = sizeof(u32),
 231         .action         = ipmr_rule_action,
 232         .match          = ipmr_rule_match,
 233         .configure      = ipmr_rule_configure,
 234         .compare        = ipmr_rule_compare,
 235         .default_pref   = fib_default_rule_pref,
 236         .fill           = ipmr_rule_fill,
 237         .nlgroup        = RTNLGRP_IPV4_RULE,
 238         .policy         = ipmr_rule_policy,
 239         .owner          = THIS_MODULE,
 240 };
 241
 242 static int __net_init ipmr_rules_init(struct net *net)
 243 {
 244         struct fib_rules_ops *ops;
 245         struct mr_table *mrt;
 246         int err;
 247
 248         ops = fib_rules_register(&ipmr_rules_ops_template, net);
 249         if (IS_ERR(ops))
 250                 return PTR_ERR(ops);
 251
 252         INIT_LIST_HEAD(&net->ipv4.mr_tables);
 253
 254         mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
 255         if (mrt == NULL) {
 256                 err = -ENOMEM;
 257                 goto err1;
 258         }
 259
 260         err = fib_default_rule_add(ops, 0x7fff, RT_TABLE_DEFAULT, 0);
 261         if (err < 0)
 262                 goto err2;
 263
 264         net->ipv4.mr_rules_ops = ops;
 265         return 0;
 266
 267 err2:
 268         kfree(mrt);
 269 err1:
 270         fib_rules_unregister(ops);
 271         return err;
 272 }
 273
 274 static void __net_exit ipmr_rules_exit(struct net *net)
 275 {
 276         struct mr_table *mrt, *next;
 277
 278         list_for_each_entry_safe(mrt, next, &net->ipv4.mr_tables, list) {
 279                 list_del(&mrt->list);
 280                 ipmr_free_table(mrt);
 281         }
 282         fib_rules_unregister(net->ipv4.mr_rules_ops);
 283 }
 284 #else
 285 #define ipmr_for_each_table(mrt, net) \
 286         for (mrt = net->ipv4.mrt; mrt; mrt = NULL)
 287
 288 static struct mr_table *ipmr_get_table(struct net *net, u32 id)
 289 {
 290         return net->ipv4.mrt;
 291 }
 292
 293 static int ipmr_fib_lookup(struct net *net, struct flowi4 *flp4,
 294                            struct mr_table **mrt)
 295 {
 296         *mrt = net->ipv4.mrt;
 297         return 0;
 298 }
 299
 300 static int __net_init ipmr_rules_init(struct net *net)
 301 {
 302         net->ipv4.mrt = ipmr_new_table(net, RT_TABLE_DEFAULT);
 303         return net->ipv4.mrt ? 0 : -ENOMEM;
 304 }
 305
 306 static void __net_exit ipmr_rules_exit(struct net *net)
 307 {
 308         ipmr_free_table(net->ipv4.mrt);
 309 }
 310 #endif
 311
 312 static struct mr_table *ipmr_new_table(struct net *net, u32 id)
 313 {
 314         struct mr_table *mrt;
 315         unsigned int i;
 316
 317         mrt = ipmr_get_table(net, id);
 318         if (mrt != NULL)
 319                 return mrt;
 320
 321         mrt = kzalloc(sizeof(*mrt), GFP_KERNEL);
 322         if (mrt == NULL)
 323                 return NULL;
 324         write_pnet(&mrt->net, net);
 325         mrt->id = id;
 326
 327         /* Forwarding cache */
 328         for (i = 0; i < MFC_LINES; i++)
 329                 INIT_LIST_HEAD(&mrt->mfc_cache_array[i]);
 330
 331         INIT_LIST_HEAD(&mrt->mfc_unres_queue);
 332
 333         setup_timer(&mrt->ipmr_expire_timer, ipmr_expire_process,
 334                     (unsigned long)mrt);
 335
 336 #ifdef CONFIG_IP_PIMSM
 337         mrt->mroute_reg_vif_num = -1;
 338 #endif
 339 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
 340         list_add_tail_rcu(&mrt->list, &net->ipv4.mr_tables);
 341 #endif
 342         return mrt;
 343 }
 344
 345 static void ipmr_free_table(struct mr_table *mrt)
 346 {
 347         del_timer_sync(&mrt->ipmr_expire_timer);
 348         mroute_clean_tables(mrt);
 349         kfree(mrt);
 350 }
 351
 352 /* Service routines creating virtual interfaces: DVMRP tunnels and PIMREG */
 353
 354 static void ipmr_del_tunnel(struct net_device *dev, struct vifctl *v)
 355 {
 356         struct net *net = dev_net(dev);
 357
 358         dev_close(dev);
 359
 360         dev = __dev_get_by_name(net, "tunl0");
 361         if (dev) {
 362                 const struct net_device_ops *ops = dev->netdev_ops;
 363                 struct ifreq ifr;
 364                 struct ip_tunnel_parm p;
 365
 366                 memset(&p, 0, sizeof(p));
 367                 p.iph.daddr = v->vifc_rmt_addr.s_addr;
 368                 p.iph.saddr = v->vifc_lcl_addr.s_addr;
 369                 p.iph.version = 4;
 370                 p.iph.ihl = 5;
 371                 p.iph.protocol = IPPROTO_IPIP;
 372                 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
 373                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
 374
 375                 if (ops->ndo_do_ioctl) {
 376                         mm_segment_t oldfs = get_fs();
 377
 378                         set_fs(KERNEL_DS);
 379                         ops->ndo_do_ioctl(dev, &ifr, SIOCDELTUNNEL);
 380                         set_fs(oldfs);
 381                 }
 382         }
 383 }
 384
 385 static
 386 struct net_device *ipmr_new_tunnel(struct net *net, struct vifctl *v)
 387 {
 388         struct net_device  *dev;
 389
 390         dev = __dev_get_by_name(net, "tunl0");
 391
 392         if (dev) {
 393                 const struct net_device_ops *ops = dev->netdev_ops;
 394                 int err;
 395                 struct ifreq ifr;
 396                 struct ip_tunnel_parm p;
 397                 struct in_device  *in_dev;
 398
 399                 memset(&p, 0, sizeof(p));
 400                 p.iph.daddr = v->vifc_rmt_addr.s_addr;
 401                 p.iph.saddr = v->vifc_lcl_addr.s_addr;
 402                 p.iph.version = 4;
 403                 p.iph.ihl = 5;
 404                 p.iph.protocol = IPPROTO_IPIP;
 405                 sprintf(p.name, "dvmrp%d", v->vifc_vifi);
 406                 ifr.ifr_ifru.ifru_data = (__force void __user *)&p;
 407
 408                 if (ops->ndo_do_ioctl) {
 409                         mm_segment_t oldfs = get_fs();
 410
 411                         set_fs(KERNEL_DS);
 412                         err = ops->ndo_do_ioctl(dev, &ifr, SIOCADDTUNNEL);
 413                         set_fs(oldfs);
 414                 } else {
 415                         err = -EOPNOTSUPP;
 416                 }
 417                 dev = NULL;
 418
 419                 if (err == 0 &&
 420                     (dev = __dev_get_by_name(net, p.name)) != NULL) {
 421                         dev->flags |= IFF_MULTICAST;
 422
 423                         in_dev = __in_dev_get_rtnl(dev);
 424                         if (in_dev == NULL)
 425                                 goto failure;
 426
 427                         ipv4_devconf_setall(in_dev);
 428                         IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
 429
 430                         if (dev_open(dev))
 431                                 goto failure;
 432                         dev_hold(dev);
 433                 }
 434         }
 435         return dev;
 436
 437 failure:
 438         /* allow the register to be completed before unregistering. */
 439         rtnl_unlock();
 440         rtnl_lock();
 441
 442         unregister_netdevice(dev);
 443         return NULL;
 444 }
 445
 446 #ifdef CONFIG_IP_PIMSM
 447
 448 static netdev_tx_t reg_vif_xmit(struct sk_buff *skb, struct net_device *dev)
 449 {
 450         struct net *net = dev_net(dev);
 451         struct mr_table *mrt;
 452         struct flowi4 fl4 = {
 453                 .flowi4_oif     = dev->ifindex,
 454                 .flowi4_iif     = skb->skb_iif,
 455                 .flowi4_mark    = skb->mark,
 456         };
 457         int err;
 458
 459         err = ipmr_fib_lookup(net, &fl4, &mrt);
 460         if (err < 0) {
 461                 kfree_skb(skb);
 462                 return err;
 463         }
 464
 465         read_lock(&mrt_lock);
 466         dev->stats.tx_bytes += skb->len;
 467         dev->stats.tx_packets++;
 468         ipmr_cache_report(mrt, skb, mrt->mroute_reg_vif_num, IGMPMSG_WHOLEPKT);
 469         read_unlock(&mrt_lock);
 470         kfree_skb(skb);
 471         return NETDEV_TX_OK;
 472 }
 473
 474 static const struct net_device_ops reg_vif_netdev_ops = {
 475         .ndo_start_xmit = reg_vif_xmit,
 476 };
 477
 478 static void reg_vif_setup(struct net_device *dev)
 479 {
 480         dev->type               = ARPHRD_PIMREG;
 481         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr) - 8;
 482         dev->flags              = IFF_NOARP;
 483         dev->netdev_ops         = &reg_vif_netdev_ops,
 484         dev->destructor         = free_netdev;
 485         dev->features           |= NETIF_F_NETNS_LOCAL;
 486 }
 487
 488 static struct net_device *ipmr_reg_vif(struct net *net, struct mr_table *mrt)
 489 {
 490         struct net_device *dev;
 491         struct in_device *in_dev;
 492         char name[IFNAMSIZ];
 493
 494         if (mrt->id == RT_TABLE_DEFAULT)
 495                 sprintf(name, "pimreg");
 496         else
 497                 sprintf(name, "pimreg%u", mrt->id);
 498
 499         dev = alloc_netdev(0, name, reg_vif_setup);
 500
 501         if (dev == NULL)
 502                 return NULL;
 503
 504         dev_net_set(dev, net);
 505
 506         if (register_netdevice(dev)) {
 507                 free_netdev(dev);
 508                 return NULL;
 509         }
 510         dev->iflink = 0;
 511
 512         rcu_read_lock();
 513         in_dev = __in_dev_get_rcu(dev);
 514         if (!in_dev) {
 515                 rcu_read_unlock();
 516                 goto failure;
 517         }
 518
 519         ipv4_devconf_setall(in_dev);
 520         IPV4_DEVCONF(in_dev->cnf, RP_FILTER) = 0;
 521         rcu_read_unlock();
 522
 523         if (dev_open(dev))
 524                 goto failure;
 525
 526         dev_hold(dev);
 527
 528         return dev;
 529
 530 failure:
 531         /* allow the register to be completed before unregistering. */
 532         rtnl_unlock();
 533         rtnl_lock();
 534
 535         unregister_netdevice(dev);
 536         return NULL;
 537 }
 538 #endif
 539
 540 /**
 541  *      vif_delete - Delete a VIF entry
 542  *      @notify: Set to 1, if the caller is a notifier_call
 543  */
 544
 545 static int vif_delete(struct mr_table *mrt, int vifi, int notify,
 546                       struct list_head *head)
 547 {
 548         struct vif_device *v;
 549         struct net_device *dev;
 550         struct in_device *in_dev;
 551
 552         if (vifi < 0 || vifi >= mrt->maxvif)
 553                 return -EADDRNOTAVAIL;
 554
 555         v = &mrt->vif_table[vifi];
 556
 557         write_lock_bh(&mrt_lock);
 558         dev = v->dev;
 559         v->dev = NULL;
 560
 561         if (!dev) {
 562                 write_unlock_bh(&mrt_lock);
 563                 return -EADDRNOTAVAIL;
 564         }
 565
 566 #ifdef CONFIG_IP_PIMSM
 567         if (vifi == mrt->mroute_reg_vif_num)
 568                 mrt->mroute_reg_vif_num = -1;
 569 #endif
 570
 571         if (vifi + 1 == mrt->maxvif) {
 572                 int tmp;
 573
 574                 for (tmp = vifi - 1; tmp >= 0; tmp--) {
 575                         if (VIF_EXISTS(mrt, tmp))
 576                                 break;
 577                 }
 578                 mrt->maxvif = tmp+1;
 579         }
 580
 581         write_unlock_bh(&mrt_lock);
 582
 583         dev_set_allmulti(dev, -1);
 584
 585         in_dev = __in_dev_get_rtnl(dev);
 586         if (in_dev) {
 587                 IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)--;
 588                 inet_netconf_notify_devconf(dev_net(dev),
 589                                             NETCONFA_MC_FORWARDING,
 590                                             dev->ifindex, &in_dev->cnf);
 591                 ip_rt_multicast_event(in_dev);
 592         }
 593
 594         if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER) && !notify)
 595                 unregister_netdevice_queue(dev, head);
 596
 597         dev_put(dev);
 598         return 0;
 599 }
 600
 601 static void ipmr_cache_free_rcu(struct rcu_head *head)
 602 {
 603         struct mfc_cache *c = container_of(head, struct mfc_cache, rcu);
 604
 605         kmem_cache_free(mrt_cachep, c);
 606 }
 607
 608 static inline void ipmr_cache_free(struct mfc_cache *c)
 609 {
 610         call_rcu(&c->rcu, ipmr_cache_free_rcu);
 611 }
 612
 613 /* Destroy an unresolved cache entry, killing queued skbs
 614  * and reporting error to netlink readers.
 615  */
 616
 617 static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
 618 {
 619         struct net *net = read_pnet(&mrt->net);
 620         struct sk_buff *skb;
 621         struct nlmsgerr *e;
 622
 623         atomic_dec(&mrt->cache_resolve_queue_len);
 624
 625         while ((skb = skb_dequeue(&c->mfc_un.unres.unresolved))) {
 626                 if (ip_hdr(skb)->version == 0) {
 627                         struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
 628                         nlh->nlmsg_type = NLMSG_ERROR;
 629                         nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
 630                         skb_trim(skb, nlh->nlmsg_len);
 631                         e = nlmsg_data(nlh);
 632                         e->error = -ETIMEDOUT;
 633                         memset(&e->msg, 0, sizeof(e->msg));
 634
 635                         rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
 636                 } else {
 637                         kfree_skb(skb);
 638                 }
 639         }
 640
 641         ipmr_cache_free(c);
 642 }
 643
 644
 645 /* Timer process for the unresolved queue. */
 646
 647 static void ipmr_expire_process(unsigned long arg)
 648 {
 649         struct mr_table *mrt = (struct mr_table *)arg;
 650         unsigned long now;
 651         unsigned long expires;
 652         struct mfc_cache *c, *next;
 653
 654         if (!spin_trylock(&mfc_unres_lock)) {
 655                 mod_timer(&mrt->ipmr_expire_timer, jiffies+HZ/10);
 656                 return;
 657         }
 658
 659         if (list_empty(&mrt->mfc_unres_queue))
 660                 goto out;
 661
 662         now = jiffies;
 663         expires = 10*HZ;
 664
 665         list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
 666                 if (time_after(c->mfc_un.unres.expires, now)) {
 667                         unsigned long interval = c->mfc_un.unres.expires - now;
 668                         if (interval < expires)
 669                                 expires = interval;
 670                         continue;
 671                 }
 672
 673                 list_del(&c->list);
 674                 mroute_netlink_event(mrt, c, RTM_DELROUTE);
 675                 ipmr_destroy_unres(mrt, c);
 676         }
 677
 678         if (!list_empty(&mrt->mfc_unres_queue))
 679                 mod_timer(&mrt->ipmr_expire_timer, jiffies + expires);
 680
 681 out:
 682         spin_unlock(&mfc_unres_lock);
 683 }
 684
 685 /* Fill oifs list. It is called under write locked mrt_lock. */
 686
 687 static void ipmr_update_thresholds(struct mr_table *mrt, struct mfc_cache *cache,
 688                                    unsigned char *ttls)
 689 {
 690         int vifi;
 691
 692         cache->mfc_un.res.minvif = MAXVIFS;
 693         cache->mfc_un.res.maxvif = 0;
 694         memset(cache->mfc_un.res.ttls, 255, MAXVIFS);
 695
 696         for (vifi = 0; vifi < mrt->maxvif; vifi++) {
 697                 if (VIF_EXISTS(mrt, vifi) &&
 698                     ttls[vifi] && ttls[vifi] < 255) {
 699                         cache->mfc_un.res.ttls[vifi] = ttls[vifi];
 700                         if (cache->mfc_un.res.minvif > vifi)
 701                                 cache->mfc_un.res.minvif = vifi;
 702                         if (cache->mfc_un.res.maxvif <= vifi)
 703                                 cache->mfc_un.res.maxvif = vifi + 1;
 704                 }
 705         }
 706 }
 707
 708 static int vif_add(struct net *net, struct mr_table *mrt,
 709                    struct vifctl *vifc, int mrtsock)
 710 {
 711         int vifi = vifc->vifc_vifi;
 712         struct vif_device *v = &mrt->vif_table[vifi];
 713         struct net_device *dev;
 714         struct in_device *in_dev;
 715         int err;
 716
 717         /* Is vif busy ? */
 718         if (VIF_EXISTS(mrt, vifi))
 719                 return -EADDRINUSE;
 720
 721         switch (vifc->vifc_flags) {
 722 #ifdef CONFIG_IP_PIMSM
 723         case VIFF_REGISTER:
 724                 /*
 725                  * Special Purpose VIF in PIM
 726                  * All the packets will be sent to the daemon
 727                  */
 728                 if (mrt->mroute_reg_vif_num >= 0)
 729                         return -EADDRINUSE;
 730                 dev = ipmr_reg_vif(net, mrt);
 731                 if (!dev)
 732                         return -ENOBUFS;
 733                 err = dev_set_allmulti(dev, 1);
 734                 if (err) {
 735                         unregister_netdevice(dev);
 736                         dev_put(dev);
 737                         return err;
 738                 }
 739                 break;
 740 #endif
 741         case VIFF_TUNNEL:
 742                 dev = ipmr_new_tunnel(net, vifc);
 743                 if (!dev)
 744                         return -ENOBUFS;
 745                 err = dev_set_allmulti(dev, 1);
 746                 if (err) {
 747                         ipmr_del_tunnel(dev, vifc);
 748                         dev_put(dev);
 749                         return err;
 750                 }
 751                 break;
 752
 753         case VIFF_USE_IFINDEX:
 754         case 0:
 755                 if (vifc->vifc_flags == VIFF_USE_IFINDEX) {
 756                         dev = dev_get_by_index(net, vifc->vifc_lcl_ifindex);
 757                         if (dev && __in_dev_get_rtnl(dev) == NULL) {
 758                                 dev_put(dev);
 759                                 return -EADDRNOTAVAIL;
 760                         }
 761                 } else {
 762                         dev = ip_dev_find(net, vifc->vifc_lcl_addr.s_addr);
 763                 }
 764                 if (!dev)
 765                         return -EADDRNOTAVAIL;
 766                 err = dev_set_allmulti(dev, 1);
 767                 if (err) {
 768                         dev_put(dev);
 769                         return err;
 770                 }
 771                 break;
 772         default:
 773                 return -EINVAL;
 774         }
 775
 776         in_dev = __in_dev_get_rtnl(dev);
 777         if (!in_dev) {
 778                 dev_put(dev);
 779                 return -EADDRNOTAVAIL;
 780         }
 781         IPV4_DEVCONF(in_dev->cnf, MC_FORWARDING)++;
 782         inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING, dev->ifindex,
 783                                     &in_dev->cnf);
 784         ip_rt_multicast_event(in_dev);
 785
 786         /* Fill in the VIF structures */
 787
 788         v->rate_limit = vifc->vifc_rate_limit;
 789         v->local = vifc->vifc_lcl_addr.s_addr;
 790         v->remote = vifc->vifc_rmt_addr.s_addr;
 791         v->flags = vifc->vifc_flags;
 792         if (!mrtsock)
 793                 v->flags |= VIFF_STATIC;
 794         v->threshold = vifc->vifc_threshold;
 795         v->bytes_in = 0;
 796         v->bytes_out = 0;
 797         v->pkt_in = 0;
 798         v->pkt_out = 0;
 799         v->link = dev->ifindex;
 800         if (v->flags & (VIFF_TUNNEL | VIFF_REGISTER))
 801                 v->link = dev->iflink;
 802
 803         /* And finish update writing critical data */
 804         write_lock_bh(&mrt_lock);
 805         v->dev = dev;
 806 #ifdef CONFIG_IP_PIMSM
 807         if (v->flags & VIFF_REGISTER)
 808                 mrt->mroute_reg_vif_num = vifi;
 809 #endif
 810         if (vifi+1 > mrt->maxvif)
 811                 mrt->maxvif = vifi+1;
 812         write_unlock_bh(&mrt_lock);
 813         return 0;
 814 }
 815
 816 /* called with rcu_read_lock() */
 817 static struct mfc_cache *ipmr_cache_find(struct mr_table *mrt,
 818                                          __be32 origin,
 819                                          __be32 mcastgrp)
 820 {
 821         int line = MFC_HASH(mcastgrp, origin);
 822         struct mfc_cache *c;
 823
 824         list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list) {
 825                 if (c->mfc_origin == origin && c->mfc_mcastgrp == mcastgrp)
 826                         return c;
 827         }
 828         return NULL;
 829 }
 830
 831 /* Look for a (*,*,oif) entry */
 832 static struct mfc_cache *ipmr_cache_find_any_parent(struct mr_table *mrt,
 833                                                     int vifi)
 834 {
 835         int line = MFC_HASH(htonl(INADDR_ANY), htonl(INADDR_ANY));
 836         struct mfc_cache *c;
 837
 838         list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list)
 839                 if (c->mfc_origin == htonl(INADDR_ANY) &&
 840                     c->mfc_mcastgrp == htonl(INADDR_ANY) &&
 841                     c->mfc_un.res.ttls[vifi] < 255)
 842                         return c;
 843
 844         return NULL;
 845 }
 846
 847 /* Look for a (*,G) entry */
 848 static struct mfc_cache *ipmr_cache_find_any(struct mr_table *mrt,
 849                                              __be32 mcastgrp, int vifi)
 850 {
 851         int line = MFC_HASH(mcastgrp, htonl(INADDR_ANY));
 852         struct mfc_cache *c, *proxy;
 853
 854         if (mcastgrp == htonl(INADDR_ANY))
 855                 goto skip;
 856
 857         list_for_each_entry_rcu(c, &mrt->mfc_cache_array[line], list)
 858                 if (c->mfc_origin == htonl(INADDR_ANY) &&
 859                     c->mfc_mcastgrp == mcastgrp) {
 860                         if (c->mfc_un.res.ttls[vifi] < 255)
 861                                 return c;
 862
 863                         /* It's ok if the vifi is part of the static tree */
 864                         proxy = ipmr_cache_find_any_parent(mrt,
 865                                                            c->mfc_parent);
 866                         if (proxy && proxy->mfc_un.res.ttls[vifi] < 255)
 867                                 return c;
 868                 }
 869
 870 skip:
 871         return ipmr_cache_find_any_parent(mrt, vifi);
 872 }
 873
 874 /*
 875  *      Allocate a multicast cache entry
 876  */
 877 static struct mfc_cache *ipmr_cache_alloc(void)
 878 {
 879         struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_KERNEL);
 880
 881         if (c)
 882                 c->mfc_un.res.minvif = MAXVIFS;
 883         return c;
 884 }
 885
 886 static struct mfc_cache *ipmr_cache_alloc_unres(void)
 887 {
 888         struct mfc_cache *c = kmem_cache_zalloc(mrt_cachep, GFP_ATOMIC);
 889
 890         if (c) {
 891                 skb_queue_head_init(&c->mfc_un.unres.unresolved);
 892                 c->mfc_un.unres.expires = jiffies + 10*HZ;
 893         }
 894         return c;
 895 }
 896
 897 /*
 898  *      A cache entry has gone into a resolved state from queued
 899  */
 900
 901 static void ipmr_cache_resolve(struct net *net, struct mr_table *mrt,
 902                                struct mfc_cache *uc, struct mfc_cache *c)
 903 {
 904         struct sk_buff *skb;
 905         struct nlmsgerr *e;
 906
 907         /* Play the pending entries through our router */
 908
 909         while ((skb = __skb_dequeue(&uc->mfc_un.unres.unresolved))) {
 910                 if (ip_hdr(skb)->version == 0) {
 911                         struct nlmsghdr *nlh = (struct nlmsghdr *)skb_pull(skb, sizeof(struct iphdr));
 912
 913                         if (__ipmr_fill_mroute(mrt, skb, c, nlmsg_data(nlh)) > 0) {
 914                                 nlh->nlmsg_len = skb_tail_pointer(skb) -
 915                                                  (u8 *)nlh;
 916                         } else {
 917                                 nlh->nlmsg_type = NLMSG_ERROR;
 918                                 nlh->nlmsg_len = nlmsg_msg_size(sizeof(struct nlmsgerr));
 919                                 skb_trim(skb, nlh->nlmsg_len);
 920                                 e = nlmsg_data(nlh);
 921                                 e->error = -EMSGSIZE;
 922                                 memset(&e->msg, 0, sizeof(e->msg));
 923                         }
 924
 925                         rtnl_unicast(skb, net, NETLINK_CB(skb).portid);
 926                 } else {
 927                         ip_mr_forward(net, mrt, skb, c, 0);
 928                 }
 929         }
 930 }
 931
 932 /*
 933  *      Bounce a cache query up to mrouted. We could use netlink for this but mrouted
 934  *      expects the following bizarre scheme.
 935  *
 936  *      Called under mrt_lock.
 937  */
 938
 939 static int ipmr_cache_report(struct mr_table *mrt,
 940                              struct sk_buff *pkt, vifi_t vifi, int assert)
 941 {
 942         struct sk_buff *skb;
 943         const int ihl = ip_hdrlen(pkt);
 944         struct igmphdr *igmp;
 945         struct igmpmsg *msg;
 946         struct sock *mroute_sk;
 947         int ret;
 948         unsigned long tail_offset;
 949
 950 #ifdef CONFIG_IP_PIMSM
 951         if (assert == IGMPMSG_WHOLEPKT)
 952                 skb = skb_realloc_headroom(pkt, sizeof(struct iphdr));
 953         else
 954 #endif
 955                 skb = alloc_skb(128, GFP_ATOMIC);
 956
 957         if (!skb)
 958                 return -ENOBUFS;
 959
 960 #ifdef CONFIG_IP_PIMSM
 961         if (assert == IGMPMSG_WHOLEPKT) {
 962                 /* Ugly, but we have no choice with this interface.
 963                  * Duplicate old header, fix ihl, length etc.
 964                  * And all this only to mangle msg->im_msgtype and
 965                  * to set msg->im_mbz to "mbz" :-)
 966                  */
 967                 skb_push(skb, sizeof(struct iphdr));
 968                 skb_reset_network_header(skb);
 969                 skb_reset_transport_header(skb);
 970                 msg = (struct igmpmsg *)skb_network_header(skb);
 971                 memcpy(msg, skb_network_header(pkt), sizeof(struct iphdr));
 972                 msg->im_msgtype = IGMPMSG_WHOLEPKT;
 973                 msg->im_mbz = 0;
 974                 msg->im_vif = mrt->mroute_reg_vif_num;
 975                 ip_hdr(skb)->ihl = sizeof(struct iphdr) >> 2;
 976                 ip_hdr(skb)->tot_len = htons(ntohs(ip_hdr(pkt)->tot_len) +
 977                                              sizeof(struct iphdr));
 978         } else
 979 #endif
 980         {
 981
 982         /* Copy the IP header */
 983
 984         tail_offset = skb_tail_offset(skb);
 985         if (tail_offset > 0xffff) {
 986                 kfree_skb(skb);
 987                 return -EINVAL;
 988         }
 989         skb_set_network_header(skb, tail_offset);
 990         skb_put(skb, ihl);
 991         skb_copy_to_linear_data(skb, pkt->data, ihl);
 992         ip_hdr(skb)->protocol = 0;      /* Flag to the kernel this is a route add */
 993         msg = (struct igmpmsg *)skb_network_header(skb);
 994         msg->im_vif = vifi;
 995         skb_dst_set(skb, dst_clone(skb_dst(pkt)));
 996
 997         /* Add our header */
 998
 999         igmp = (struct igmphdr *)skb_put(skb, sizeof(struct igmphdr));
1000         igmp->type      =
1001         msg->im_msgtype = assert;
1002         igmp->code      = 0;
1003         ip_hdr(skb)->tot_len = htons(skb->len);         /* Fix the length */
1004         skb->transport_header = skb->network_header;
1005         }
1006
1007         rcu_read_lock();
1008         mroute_sk = rcu_dereference(mrt->mroute_sk);
1009         if (mroute_sk == NULL) {
1010                 rcu_read_unlock();
1011                 kfree_skb(skb);
1012                 return -EINVAL;
1013         }
1014
1015         /* Deliver to mrouted */
1016
1017         ret = sock_queue_rcv_skb(mroute_sk, skb);
1018         rcu_read_unlock();
1019         if (ret < 0) {
1020                 net_warn_ratelimited("mroute: pending queue full, dropping entries\n");
1021                 kfree_skb(skb);
1022         }
1023
1024         return ret;
1025 }
1026
1027 /*
1028  *      Queue a packet for resolution. It gets locked cache entry!
1029  */
1030
1031 static int
1032 ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi, struct sk_buff *skb)
1033 {
1034         bool found = false;
1035         int err;
1036         struct mfc_cache *c;
1037         const struct iphdr *iph = ip_hdr(skb);
1038
1039         spin_lock_bh(&mfc_unres_lock);
1040         list_for_each_entry(c, &mrt->mfc_unres_queue, list) {
1041                 if (c->mfc_mcastgrp == iph->daddr &&
1042                     c->mfc_origin == iph->saddr) {
1043                         found = true;
1044                         break;
1045                 }
1046         }
1047
1048         if (!found) {
1049                 /* Create a new entry if allowable */
1050
1051                 if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
1052                     (c = ipmr_cache_alloc_unres()) == NULL) {
1053                         spin_unlock_bh(&mfc_unres_lock);
1054
1055                         kfree_skb(skb);
1056                         return -ENOBUFS;
1057                 }
1058
1059                 /* Fill in the new cache entry */
1060
1061                 c->mfc_parent   = -1;
1062                 c->mfc_origin   = iph->saddr;
1063                 c->mfc_mcastgrp = iph->daddr;
1064
1065                 /* Reflect first query at mrouted. */
1066
1067                 err = ipmr_cache_report(mrt, skb, vifi, IGMPMSG_NOCACHE);
1068                 if (err < 0) {
1069                         /* If the report failed throw the cache entry
1070                            out - Brad Parker
1071                          */
1072                         spin_unlock_bh(&mfc_unres_lock);
1073
1074                         ipmr_cache_free(c);
1075                         kfree_skb(skb);
1076                         return err;
1077                 }
1078
1079                 atomic_inc(&mrt->cache_resolve_queue_len);
1080                 list_add(&c->list, &mrt->mfc_unres_queue);
1081                 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1082
1083                 if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
1084                         mod_timer(&mrt->ipmr_expire_timer, c->mfc_un.unres.expires);
1085         }
1086
1087         /* See if we can append the packet */
1088
1089         if (c->mfc_un.unres.unresolved.qlen > 3) {
1090                 kfree_skb(skb);
1091                 err = -ENOBUFS;
1092         } else {
1093                 skb_queue_tail(&c->mfc_un.unres.unresolved, skb);
1094                 err = 0;
1095         }
1096
1097         spin_unlock_bh(&mfc_unres_lock);
1098         return err;
1099 }
1100
1101 /*
1102  *      MFC cache manipulation by user space mroute daemon
1103  */
1104
1105 static int ipmr_mfc_delete(struct mr_table *mrt, struct mfcctl *mfc, int parent)
1106 {
1107         int line;
1108         struct mfc_cache *c, *next;
1109
1110         line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1111
1112         list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[line], list) {
1113                 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1114                     c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr &&
1115                     (parent == -1 || parent == c->mfc_parent)) {
1116                         list_del_rcu(&c->list);
1117                         mroute_netlink_event(mrt, c, RTM_DELROUTE);
1118                         ipmr_cache_free(c);
1119                         return 0;
1120                 }
1121         }
1122         return -ENOENT;
1123 }
1124
1125 static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
1126                         struct mfcctl *mfc, int mrtsock, int parent)
1127 {
1128         bool found = false;
1129         int line;
1130         struct mfc_cache *uc, *c;
1131
1132         if (mfc->mfcc_parent >= MAXVIFS)
1133                 return -ENFILE;
1134
1135         line = MFC_HASH(mfc->mfcc_mcastgrp.s_addr, mfc->mfcc_origin.s_addr);
1136
1137         list_for_each_entry(c, &mrt->mfc_cache_array[line], list) {
1138                 if (c->mfc_origin == mfc->mfcc_origin.s_addr &&
1139                     c->mfc_mcastgrp == mfc->mfcc_mcastgrp.s_addr &&
1140                     (parent == -1 || parent == c->mfc_parent)) {
1141                         found = true;
1142                         break;
1143                 }
1144         }
1145
1146         if (found) {
1147                 write_lock_bh(&mrt_lock);
1148                 c->mfc_parent = mfc->mfcc_parent;
1149                 ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1150                 if (!mrtsock)
1151                         c->mfc_flags |= MFC_STATIC;
1152                 write_unlock_bh(&mrt_lock);
1153                 mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1154                 return 0;
1155         }
1156
1157         if (mfc->mfcc_mcastgrp.s_addr != htonl(INADDR_ANY) &&
1158             !ipv4_is_multicast(mfc->mfcc_mcastgrp.s_addr))
1159                 return -EINVAL;
1160
1161         c = ipmr_cache_alloc();
1162         if (c == NULL)
1163                 return -ENOMEM;
1164
1165         c->mfc_origin = mfc->mfcc_origin.s_addr;
1166         c->mfc_mcastgrp = mfc->mfcc_mcastgrp.s_addr;
1167         c->mfc_parent = mfc->mfcc_parent;
1168         ipmr_update_thresholds(mrt, c, mfc->mfcc_ttls);
1169         if (!mrtsock)
1170                 c->mfc_flags |= MFC_STATIC;
1171
1172         list_add_rcu(&c->list, &mrt->mfc_cache_array[line]);
1173
1174         /*
1175          *      Check to see if we resolved a queued list. If so we
1176          *      need to send on the frames and tidy up.
1177          */
1178         found = false;
1179         spin_lock_bh(&mfc_unres_lock);
1180         list_for_each_entry(uc, &mrt->mfc_unres_queue, list) {
1181                 if (uc->mfc_origin == c->mfc_origin &&
1182                     uc->mfc_mcastgrp == c->mfc_mcastgrp) {
1183                         list_del(&uc->list);
1184                         atomic_dec(&mrt->cache_resolve_queue_len);
1185                         found = true;
1186                         break;
1187                 }
1188         }
1189         if (list_empty(&mrt->mfc_unres_queue))
1190                 del_timer(&mrt->ipmr_expire_timer);
1191         spin_unlock_bh(&mfc_unres_lock);
1192
1193         if (found) {
1194                 ipmr_cache_resolve(net, mrt, uc, c);
1195                 ipmr_cache_free(uc);
1196         }
1197         mroute_netlink_event(mrt, c, RTM_NEWROUTE);
1198         return 0;
1199 }
1200
1201 /*
1202  *      Close the multicast socket, and clear the vif tables etc
1203  */
1204
1205 static void mroute_clean_tables(struct mr_table *mrt)
1206 {
1207         int i;
1208         LIST_HEAD(list);
1209         struct mfc_cache *c, *next;
1210
1211         /* Shut down all active vif entries */
1212
1213         for (i = 0; i < mrt->maxvif; i++) {
1214                 if (!(mrt->vif_table[i].flags & VIFF_STATIC))
1215                         vif_delete(mrt, i, 0, &list);
1216         }
1217         unregister_netdevice_many(&list);
1218
1219         /* Wipe the cache */
1220
1221         for (i = 0; i < MFC_LINES; i++) {
1222                 list_for_each_entry_safe(c, next, &mrt->mfc_cache_array[i], list) {
1223                         if (c->mfc_flags & MFC_STATIC)
1224                                 continue;
1225                         list_del_rcu(&c->list);
1226                         mroute_netlink_event(mrt, c, RTM_DELROUTE);
1227                         ipmr_cache_free(c);
1228                 }
1229         }
1230
1231         if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
1232                 spin_lock_bh(&mfc_unres_lock);
1233                 list_for_each_entry_safe(c, next, &mrt->mfc_unres_queue, list) {
1234                         list_del(&c->list);
1235                         mroute_netlink_event(mrt, c, RTM_DELROUTE);
1236                         ipmr_destroy_unres(mrt, c);
1237                 }
1238                 spin_unlock_bh(&mfc_unres_lock);
1239         }
1240 }
1241
1242 /* called from ip_ra_control(), before an RCU grace period,
1243  * we dont need to call synchronize_rcu() here
1244  */
1245 static void mrtsock_destruct(struct sock *sk)
1246 {
1247         struct net *net = sock_net(sk);
1248         struct mr_table *mrt;
1249
1250         rtnl_lock();
1251         ipmr_for_each_table(mrt, net) {
1252                 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1253                         IPV4_DEVCONF_ALL(net, MC_FORWARDING)--;
1254                         inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1255                                                     NETCONFA_IFINDEX_ALL,
1256                                                     net->ipv4.devconf_all);
1257                         RCU_INIT_POINTER(mrt->mroute_sk, NULL);
1258                         mroute_clean_tables(mrt);
1259                 }
1260         }
1261         rtnl_unlock();
1262 }
1263
1264 /*
1265  *      Socket options and virtual interface manipulation. The whole
1266  *      virtual interface system is a complete heap, but unfortunately
1267  *      that's how BSD mrouted happens to think. Maybe one day with a proper
1268  *      MOSPF/PIM router set up we can clean this up.
1269  */
1270
1271 int ip_mroute_setsockopt(struct sock *sk, int optname, char __user *optval, unsigned int optlen)
1272 {
1273         int ret, parent = 0;
1274         struct vifctl vif;
1275         struct mfcctl mfc;
1276         struct net *net = sock_net(sk);
1277         struct mr_table *mrt;
1278
1279         if (sk->sk_type != SOCK_RAW ||
1280             inet_sk(sk)->inet_num != IPPROTO_IGMP)
1281                 return -EOPNOTSUPP;
1282
1283         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1284         if (mrt == NULL)
1285                 return -ENOENT;
1286
1287         if (optname != MRT_INIT) {
1288                 if (sk != rcu_access_pointer(mrt->mroute_sk) &&
1289                     !ns_capable(net->user_ns, CAP_NET_ADMIN))
1290                         return -EACCES;
1291         }
1292
1293         switch (optname) {
1294         case MRT_INIT:
1295                 if (optlen != sizeof(int))
1296                         return -EINVAL;
1297
1298                 rtnl_lock();
1299                 if (rtnl_dereference(mrt->mroute_sk)) {
1300                         rtnl_unlock();
1301                         return -EADDRINUSE;
1302                 }
1303
1304                 ret = ip_ra_control(sk, 1, mrtsock_destruct);
1305                 if (ret == 0) {
1306                         rcu_assign_pointer(mrt->mroute_sk, sk);
1307                         IPV4_DEVCONF_ALL(net, MC_FORWARDING)++;
1308                         inet_netconf_notify_devconf(net, NETCONFA_MC_FORWARDING,
1309                                                     NETCONFA_IFINDEX_ALL,
1310                                                     net->ipv4.devconf_all);
1311                 }
1312                 rtnl_unlock();
1313                 return ret;
1314         case MRT_DONE:
1315                 if (sk != rcu_access_pointer(mrt->mroute_sk))
1316                         return -EACCES;
1317                 return ip_ra_control(sk, 0, NULL);
1318         case MRT_ADD_VIF:
1319         case MRT_DEL_VIF:
1320                 if (optlen != sizeof(vif))
1321                         return -EINVAL;
1322                 if (copy_from_user(&vif, optval, sizeof(vif)))
1323                         return -EFAULT;
1324                 if (vif.vifc_vifi >= MAXVIFS)
1325                         return -ENFILE;
1326                 rtnl_lock();
1327                 if (optname == MRT_ADD_VIF) {
1328                         ret = vif_add(net, mrt, &vif,
1329                                       sk == rtnl_dereference(mrt->mroute_sk));
1330                 } else {
1331                         ret = vif_delete(mrt, vif.vifc_vifi, 0, NULL);
1332                 }
1333                 rtnl_unlock();
1334                 return ret;
1335
1336                 /*
1337                  *      Manipulate the forwarding caches. These live
1338                  *      in a sort of kernel/user symbiosis.
1339                  */
1340         case MRT_ADD_MFC:
1341         case MRT_DEL_MFC:
1342                 parent = -1;
1343         case MRT_ADD_MFC_PROXY:
1344         case MRT_DEL_MFC_PROXY:
1345                 if (optlen != sizeof(mfc))
1346                         return -EINVAL;
1347                 if (copy_from_user(&mfc, optval, sizeof(mfc)))
1348                         return -EFAULT;
1349                 if (parent == 0)
1350                         parent = mfc.mfcc_parent;
1351                 rtnl_lock();
1352                 if (optname == MRT_DEL_MFC || optname == MRT_DEL_MFC_PROXY)
1353                         ret = ipmr_mfc_delete(mrt, &mfc, parent);
1354                 else
1355                         ret = ipmr_mfc_add(net, mrt, &mfc,
1356                                            sk == rtnl_dereference(mrt->mroute_sk),
1357                                            parent);
1358                 rtnl_unlock();
1359                 return ret;
1360                 /*
1361                  *      Control PIM assert.
1362                  */
1363         case MRT_ASSERT:
1364         {
1365                 int v;
1366                 if (optlen != sizeof(v))
1367                         return -EINVAL;
1368                 if (get_user(v, (int __user *)optval))
1369                         return -EFAULT;
1370                 mrt->mroute_do_assert = v;
1371                 return 0;
1372         }
1373 #ifdef CONFIG_IP_PIMSM
1374         case MRT_PIM:
1375         {
1376                 int v;
1377
1378                 if (optlen != sizeof(v))
1379                         return -EINVAL;
1380                 if (get_user(v, (int __user *)optval))
1381                         return -EFAULT;
1382                 v = !!v;
1383
1384                 rtnl_lock();
1385                 ret = 0;
1386                 if (v != mrt->mroute_do_pim) {
1387                         mrt->mroute_do_pim = v;
1388                         mrt->mroute_do_assert = v;
1389                 }
1390                 rtnl_unlock();
1391                 return ret;
1392         }
1393 #endif
1394 #ifdef CONFIG_IP_MROUTE_MULTIPLE_TABLES
1395         case MRT_TABLE:
1396         {
1397                 u32 v;
1398
1399                 if (optlen != sizeof(u32))
1400                         return -EINVAL;
1401                 if (get_user(v, (u32 __user *)optval))
1402                         return -EFAULT;
1403
1404                 /* "pimreg%u" should not exceed 16 bytes (IFNAMSIZ) */
1405                 if (v != RT_TABLE_DEFAULT && v >= 1000000000)
1406                         return -EINVAL;
1407
1408                 rtnl_lock();
1409                 ret = 0;
1410                 if (sk == rtnl_dereference(mrt->mroute_sk)) {
1411                         ret = -EBUSY;
1412                 } else {
1413                         if (!ipmr_new_table(net, v))
1414                                 ret = -ENOMEM;
1415                         else
1416                                 raw_sk(sk)->ipmr_table = v;
1417                 }
1418                 rtnl_unlock();
1419                 return ret;
1420         }
1421 #endif
1422         /*
1423          *      Spurious command, or MRT_VERSION which you cannot
1424          *      set.
1425          */
1426         default:
1427                 return -ENOPROTOOPT;
1428         }
1429 }
1430
1431 /*
1432  *      Getsock opt support for the multicast routing system.
1433  */
1434
1435 int ip_mroute_getsockopt(struct sock *sk, int optname, char __user *optval, int __user *optlen)
1436 {
1437         int olr;
1438         int val;
1439         struct net *net = sock_net(sk);
1440         struct mr_table *mrt;
1441
1442         if (sk->sk_type != SOCK_RAW ||
1443             inet_sk(sk)->inet_num != IPPROTO_IGMP)
1444                 return -EOPNOTSUPP;
1445
1446         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1447         if (mrt == NULL)
1448                 return -ENOENT;
1449
1450         if (optname != MRT_VERSION &&
1451 #ifdef CONFIG_IP_PIMSM
1452            optname != MRT_PIM &&
1453 #endif
1454            optname != MRT_ASSERT)
1455                 return -ENOPROTOOPT;
1456
1457         if (get_user(olr, optlen))
1458                 return -EFAULT;
1459
1460         olr = min_t(unsigned int, olr, sizeof(int));
1461         if (olr < 0)
1462                 return -EINVAL;
1463
1464         if (put_user(olr, optlen))
1465                 return -EFAULT;
1466         if (optname == MRT_VERSION)
1467                 val = 0x0305;
1468 #ifdef CONFIG_IP_PIMSM
1469         else if (optname == MRT_PIM)
1470                 val = mrt->mroute_do_pim;
1471 #endif
1472         else
1473                 val = mrt->mroute_do_assert;
1474         if (copy_to_user(optval, &val, olr))
1475                 return -EFAULT;
1476         return 0;
1477 }
1478
1479 /*
1480  *      The IP multicast ioctl support routines.
1481  */
1482
1483 int ipmr_ioctl(struct sock *sk, int cmd, void __user *arg)
1484 {
1485         struct sioc_sg_req sr;
1486         struct sioc_vif_req vr;
1487         struct vif_device *vif;
1488         struct mfc_cache *c;
1489         struct net *net = sock_net(sk);
1490         struct mr_table *mrt;
1491
1492         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1493         if (mrt == NULL)
1494                 return -ENOENT;
1495
1496         switch (cmd) {
1497         case SIOCGETVIFCNT:
1498                 if (copy_from_user(&vr, arg, sizeof(vr)))
1499                         return -EFAULT;
1500                 if (vr.vifi >= mrt->maxvif)
1501                         return -EINVAL;
1502                 read_lock(&mrt_lock);
1503                 vif = &mrt->vif_table[vr.vifi];
1504                 if (VIF_EXISTS(mrt, vr.vifi)) {
1505                         vr.icount = vif->pkt_in;
1506                         vr.ocount = vif->pkt_out;
1507                         vr.ibytes = vif->bytes_in;
1508                         vr.obytes = vif->bytes_out;
1509                         read_unlock(&mrt_lock);
1510
1511                         if (copy_to_user(arg, &vr, sizeof(vr)))
1512                                 return -EFAULT;
1513                         return 0;
1514                 }
1515                 read_unlock(&mrt_lock);
1516                 return -EADDRNOTAVAIL;
1517         case SIOCGETSGCNT:
1518                 if (copy_from_user(&sr, arg, sizeof(sr)))
1519                         return -EFAULT;
1520
1521                 rcu_read_lock();
1522                 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1523                 if (c) {
1524                         sr.pktcnt = c->mfc_un.res.pkt;
1525                         sr.bytecnt = c->mfc_un.res.bytes;
1526                         sr.wrong_if = c->mfc_un.res.wrong_if;
1527                         rcu_read_unlock();
1528
1529                         if (copy_to_user(arg, &sr, sizeof(sr)))
1530                                 return -EFAULT;
1531                         return 0;
1532                 }
1533                 rcu_read_unlock();
1534                 return -EADDRNOTAVAIL;
1535         default:
1536                 return -ENOIOCTLCMD;
1537         }
1538 }
1539
1540 #ifdef CONFIG_COMPAT
1541 struct compat_sioc_sg_req {
1542         struct in_addr src;
1543         struct in_addr grp;
1544         compat_ulong_t pktcnt;
1545         compat_ulong_t bytecnt;
1546         compat_ulong_t wrong_if;
1547 };
1548
1549 struct compat_sioc_vif_req {
1550         vifi_t  vifi;           /* Which iface */
1551         compat_ulong_t icount;
1552         compat_ulong_t ocount;
1553         compat_ulong_t ibytes;
1554         compat_ulong_t obytes;
1555 };
1556
1557 int ipmr_compat_ioctl(struct sock *sk, unsigned int cmd, void __user *arg)
1558 {
1559         struct compat_sioc_sg_req sr;
1560         struct compat_sioc_vif_req vr;
1561         struct vif_device *vif;
1562         struct mfc_cache *c;
1563         struct net *net = sock_net(sk);
1564         struct mr_table *mrt;
1565
1566         mrt = ipmr_get_table(net, raw_sk(sk)->ipmr_table ? : RT_TABLE_DEFAULT);
1567         if (mrt == NULL)
1568                 return -ENOENT;
1569
1570         switch (cmd) {
1571         case SIOCGETVIFCNT:
1572                 if (copy_from_user(&vr, arg, sizeof(vr)))
1573                         return -EFAULT;
1574                 if (vr.vifi >= mrt->maxvif)
1575                         return -EINVAL;
1576                 read_lock(&mrt_lock);
1577                 vif = &mrt->vif_table[vr.vifi];
1578                 if (VIF_EXISTS(mrt, vr.vifi)) {
1579                         vr.icount = vif->pkt_in;
1580                         vr.ocount = vif->pkt_out;
1581                         vr.ibytes = vif->bytes_in;
1582                         vr.obytes = vif->bytes_out;
1583                         read_unlock(&mrt_lock);
1584
1585                         if (copy_to_user(arg, &vr, sizeof(vr)))
1586                                 return -EFAULT;
1587                         return 0;
1588                 }
1589                 read_unlock(&mrt_lock);
1590                 return -EADDRNOTAVAIL;
1591         case SIOCGETSGCNT:
1592                 if (copy_from_user(&sr, arg, sizeof(sr)))
1593                         return -EFAULT;
1594
1595                 rcu_read_lock();
1596                 c = ipmr_cache_find(mrt, sr.src.s_addr, sr.grp.s_addr);
1597                 if (c) {
1598                         sr.pktcnt = c->mfc_un.res.pkt;
1599                         sr.bytecnt = c->mfc_un.res.bytes;
1600                         sr.wrong_if = c->mfc_un.res.wrong_if;
1601                         rcu_read_unlock();
1602
1603                         if (copy_to_user(arg, &sr, sizeof(sr)))
1604                                 return -EFAULT;
1605                         return 0;
1606                 }
1607                 rcu_read_unlock();
1608                 return -EADDRNOTAVAIL;
1609         default:
1610                 return -ENOIOCTLCMD;
1611         }
1612 }
1613 #endif
1614
1615
1616 static int ipmr_device_event(struct notifier_block *this, unsigned long event, void *ptr)
1617 {
1618         struct net_device *dev = netdev_notifier_info_to_dev(ptr);
1619         struct net *net = dev_net(dev);
1620         struct mr_table *mrt;
1621         struct vif_device *v;
1622         int ct;
1623
1624         if (event != NETDEV_UNREGISTER)
1625                 return NOTIFY_DONE;
1626
1627         ipmr_for_each_table(mrt, net) {
1628                 v = &mrt->vif_table[0];
1629                 for (ct = 0; ct < mrt->maxvif; ct++, v++) {
1630                         if (v->dev == dev)
1631                                 vif_delete(mrt, ct, 1, NULL);
1632                 }
1633         }
1634         return NOTIFY_DONE;
1635 }
1636
1637
1638 static struct notifier_block ip_mr_notifier = {
1639         .notifier_call = ipmr_device_event,
1640 };
1641
1642 /*
1643  *      Encapsulate a packet by attaching a valid IPIP header to it.
1644  *      This avoids tunnel drivers and other mess and gives us the speed so
1645  *      important for multicast video.
1646  */
1647
1648 static void ip_encap(struct sk_buff *skb, __be32 saddr, __be32 daddr)
1649 {
1650         struct iphdr *iph;
1651         const struct iphdr *old_iph = ip_hdr(skb);
1652
1653         skb_push(skb, sizeof(struct iphdr));
1654         skb->transport_header = skb->network_header;
1655         skb_reset_network_header(skb);
1656         iph = ip_hdr(skb);
1657
1658         iph->version    =       4;
1659         iph->tos        =       old_iph->tos;
1660         iph->ttl        =       old_iph->ttl;
1661         iph->frag_off   =       0;
1662         iph->daddr      =       daddr;
1663         iph->saddr      =       saddr;
1664         iph->protocol   =       IPPROTO_IPIP;
1665         iph->ihl        =       5;
1666         iph->tot_len    =       htons(skb->len);
1667         ip_select_ident(iph, skb_dst(skb), NULL);
1668         ip_send_check(iph);
1669
1670         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
1671         nf_reset(skb);
1672 }
1673
1674 static inline int ipmr_forward_finish(struct sk_buff *skb)
1675 {
1676         struct ip_options *opt = &(IPCB(skb)->opt);
1677
1678         IP_INC_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTFORWDATAGRAMS);
1679         IP_ADD_STATS_BH(dev_net(skb_dst(skb)->dev), IPSTATS_MIB_OUTOCTETS, skb->len);
1680
1681         if (unlikely(opt->optlen))
1682                 ip_forward_options(skb);
1683
1684         return dst_output(skb);
1685 }
1686
1687 /*
1688  *      Processing handlers for ipmr_forward
1689  */
1690
1691 static void ipmr_queue_xmit(struct net *net, struct mr_table *mrt,
1692                             struct sk_buff *skb, struct mfc_cache *c, int vifi)
1693 {
1694         const struct iphdr *iph = ip_hdr(skb);
1695         struct vif_device *vif = &mrt->vif_table[vifi];
1696         struct net_device *dev;
1697         struct rtable *rt;
1698         struct flowi4 fl4;
1699         int    encap = 0;
1700
1701         if (vif->dev == NULL)
1702                 goto out_free;
1703
1704 #ifdef CONFIG_IP_PIMSM
1705         if (vif->flags & VIFF_REGISTER) {
1706                 vif->pkt_out++;
1707                 vif->bytes_out += skb->len;
1708                 vif->dev->stats.tx_bytes += skb->len;
1709                 vif->dev->stats.tx_packets++;
1710                 ipmr_cache_report(mrt, skb, vifi, IGMPMSG_WHOLEPKT);
1711                 goto out_free;
1712         }
1713 #endif
1714
1715         if (vif->flags & VIFF_TUNNEL) {
1716                 rt = ip_route_output_ports(net, &fl4, NULL,
1717                                            vif->remote, vif->local,
1718                                            0, 0,
1719                                            IPPROTO_IPIP,
1720                                            RT_TOS(iph->tos), vif->link);
1721                 if (IS_ERR(rt))
1722                         goto out_free;
1723                 encap = sizeof(struct iphdr);
1724         } else {
1725                 rt = ip_route_output_ports(net, &fl4, NULL, iph->daddr, 0,
1726                                            0, 0,
1727                                            IPPROTO_IPIP,
1728                                            RT_TOS(iph->tos), vif->link);
1729                 if (IS_ERR(rt))
1730                         goto out_free;
1731         }
1732
1733         dev = rt->dst.dev;
1734
1735         if (skb->len+encap > dst_mtu(&rt->dst) && (ntohs(iph->frag_off) & IP_DF)) {
1736                 /* Do not fragment multicasts. Alas, IPv4 does not
1737                  * allow to send ICMP, so that packets will disappear
1738                  * to blackhole.
1739                  */
1740
1741                 IP_INC_STATS_BH(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
1742                 ip_rt_put(rt);
1743                 goto out_free;
1744         }
1745
1746         encap += LL_RESERVED_SPACE(dev) + rt->dst.header_len;
1747
1748         if (skb_cow(skb, encap)) {
1749                 ip_rt_put(rt);
1750                 goto out_free;
1751         }
1752
1753         vif->pkt_out++;
1754         vif->bytes_out += skb->len;
1755
1756         skb_dst_drop(skb);
1757         skb_dst_set(skb, &rt->dst);
1758         ip_decrease_ttl(ip_hdr(skb));
1759
1760         /* FIXME: forward and output firewalls used to be called here.
1761          * What do we do with netfilter? -- RR
1762          */
1763         if (vif->flags & VIFF_TUNNEL) {
1764                 ip_encap(skb, vif->local, vif->remote);
1765                 /* FIXME: extra output firewall step used to be here. --RR */
1766                 vif->dev->stats.tx_packets++;
1767                 vif->dev->stats.tx_bytes += skb->len;
1768         }
1769
1770         IPCB(skb)->flags |= IPSKB_FORWARDED;
1771
1772         /*
1773          * RFC1584 teaches, that DVMRP/PIM router must deliver packets locally
1774          * not only before forwarding, but after forwarding on all output
1775          * interfaces. It is clear, if mrouter runs a multicasting
1776          * program, it should receive packets not depending to what interface
1777          * program is joined.
1778          * If we will not make it, the program will have to join on all
1779          * interfaces. On the other hand, multihoming host (or router, but
1780          * not mrouter) cannot join to more than one interface - it will
1781          * result in receiving multiple packets.
1782          */
1783         NF_HOOK(NFPROTO_IPV4, NF_INET_FORWARD, skb, skb->dev, dev,
1784                 ipmr_forward_finish);
1785         return;
1786
1787 out_free:
1788         kfree_skb(skb);
1789 }
1790
1791 static int ipmr_find_vif(struct mr_table *mrt, struct net_device *dev)
1792 {
1793         int ct;
1794
1795         for (ct = mrt->maxvif-1; ct >= 0; ct--) {
1796                 if (mrt->vif_table[ct].dev == dev)
1797                         break;
1798         }
1799         return ct;
1800 }
1801
1802 /* "local" means that we should preserve one skb (for local delivery) */
1803
1804 static int ip_mr_forward(struct net *net, struct mr_table *mrt,
1805                          struct sk_buff *skb, struct mfc_cache *cache,
1806                          int local)
1807 {
1808         int psend = -1;
1809         int vif, ct;
1810         int true_vifi = ipmr_find_vif(mrt, skb->dev);
1811
1812         vif = cache->mfc_parent;
1813         cache->mfc_un.res.pkt++;
1814         cache->mfc_un.res.bytes += skb->len;
1815
1816         if (cache->mfc_origin == htonl(INADDR_ANY) && true_vifi >= 0) {
1817                 struct mfc_cache *cache_proxy;
1818
1819                 /* For an (*,G) entry, we only check that the incomming
1820                  * interface is part of the static tree.
1821                  */
1822                 cache_proxy = ipmr_cache_find_any_parent(mrt, vif);
1823                 if (cache_proxy &&
1824                     cache_proxy->mfc_un.res.ttls[true_vifi] < 255)
1825                         goto forward;
1826         }
1827
1828         /*
1829          * Wrong interface: drop packet and (maybe) send PIM assert.
1830          */
1831         if (mrt->vif_table[vif].dev != skb->dev) {
1832                 if (rt_is_output_route(skb_rtable(skb))) {
1833                         /* It is our own packet, looped back.
1834                          * Very complicated situation...
1835                          *
1836                          * The best workaround until routing daemons will be
1837                          * fixed is not to redistribute packet, if it was
1838                          * send through wrong interface. It means, that
1839                          * multicast applications WILL NOT work for
1840                          * (S,G), which have default multicast route pointing
1841                          * to wrong oif. In any case, it is not a good
1842                          * idea to use multicasting applications on router.
1843                          */
1844                         goto dont_forward;
1845                 }
1846
1847                 cache->mfc_un.res.wrong_if++;
1848
1849                 if (true_vifi >= 0 && mrt->mroute_do_assert &&
1850                     /* pimsm uses asserts, when switching from RPT to SPT,
1851                      * so that we cannot check that packet arrived on an oif.
1852                      * It is bad, but otherwise we would need to move pretty
1853                      * large chunk of pimd to kernel. Ough... --ANK
1854                      */
1855                     (mrt->mroute_do_pim ||
1856                      cache->mfc_un.res.ttls[true_vifi] < 255) &&
1857                     time_after(jiffies,
1858                                cache->mfc_un.res.last_assert + MFC_ASSERT_THRESH)) {
1859                         cache->mfc_un.res.last_assert = jiffies;
1860                         ipmr_cache_report(mrt, skb, true_vifi, IGMPMSG_WRONGVIF);
1861                 }
1862                 goto dont_forward;
1863         }
1864
1865 forward:
1866         mrt->vif_table[vif].pkt_in++;
1867         mrt->vif_table[vif].bytes_in += skb->len;
1868
1869         /*
1870          *      Forward the frame
1871          */
1872         if (cache->mfc_origin == htonl(INADDR_ANY) &&
1873             cache->mfc_mcastgrp == htonl(INADDR_ANY)) {
1874                 if (true_vifi >= 0 &&
1875                     true_vifi != cache->mfc_parent &&
1876                     ip_hdr(skb)->ttl >
1877                                 cache->mfc_un.res.ttls[cache->mfc_parent]) {
1878                         /* It's an (*,*) entry and the packet is not coming from
1879                          * the upstream: forward the packet to the upstream
1880                          * only.
1881                          */
1882                         psend = cache->mfc_parent;
1883                         goto last_forward;
1884                 }
1885                 goto dont_forward;
1886         }
1887         for (ct = cache->mfc_un.res.maxvif - 1;
1888              ct >= cache->mfc_un.res.minvif; ct--) {
1889                 /* For (*,G) entry, don't forward to the incoming interface */
1890                 if ((cache->mfc_origin != htonl(INADDR_ANY) ||
1891                      ct != true_vifi) &&
1892                     ip_hdr(skb)->ttl > cache->mfc_un.res.ttls[ct]) {
1893                         if (psend != -1) {
1894                                 struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1895
1896                                 if (skb2)
1897                                         ipmr_queue_xmit(net, mrt, skb2, cache,
1898                                                         psend);
1899                         }
1900                         psend = ct;
1901                 }
1902         }
1903 last_forward:
1904         if (psend != -1) {
1905                 if (local) {
1906                         struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
1907
1908                         if (skb2)
1909                                 ipmr_queue_xmit(net, mrt, skb2, cache, psend);
1910                 } else {
1911                         ipmr_queue_xmit(net, mrt, skb, cache, psend);
1912                         return 0;
1913                 }
1914         }
1915
1916 dont_forward:
1917         if (!local)
1918                 kfree_skb(skb);
1919         return 0;
1920 }
1921
1922 static struct mr_table *ipmr_rt_fib_lookup(struct net *net, struct sk_buff *skb)
1923 {
1924         struct rtable *rt = skb_rtable(skb);
1925         struct iphdr *iph = ip_hdr(skb);
1926         struct flowi4 fl4 = {
1927                 .daddr = iph->daddr,
1928                 .saddr = iph->saddr,
1929                 .flowi4_tos = RT_TOS(iph->tos),
1930                 .flowi4_oif = (rt_is_output_route(rt) ?
1931                                skb->dev->ifindex : 0),
1932                 .flowi4_iif = (rt_is_output_route(rt) ?
1933                                LOOPBACK_IFINDEX :
1934                                skb->dev->ifindex),
1935                 .flowi4_mark = skb->mark,
1936         };
1937         struct mr_table *mrt;
1938         int err;
1939
1940         err = ipmr_fib_lookup(net, &fl4, &mrt);
1941         if (err)
1942                 return ERR_PTR(err);
1943         return mrt;
1944 }
1945
1946 /*
1947  *      Multicast packets for forwarding arrive here
1948  *      Called with rcu_read_lock();
1949  */
1950
1951 int ip_mr_input(struct sk_buff *skb)
1952 {
1953         struct mfc_cache *cache;
1954         struct net *net = dev_net(skb->dev);
1955         int local = skb_rtable(skb)->rt_flags & RTCF_LOCAL;
1956         struct mr_table *mrt;
1957
1958         /* Packet is looped back after forward, it should not be
1959          * forwarded second time, but still can be delivered locally.
1960          */
1961         if (IPCB(skb)->flags & IPSKB_FORWARDED)
1962                 goto dont_forward;
1963
1964         mrt = ipmr_rt_fib_lookup(net, skb);
1965         if (IS_ERR(mrt)) {
1966                 kfree_skb(skb);
1967                 return PTR_ERR(mrt);
1968         }
1969         if (!local) {
1970                 if (IPCB(skb)->opt.router_alert) {
1971                         if (ip_call_ra_chain(skb))
1972                                 return 0;
1973                 } else if (ip_hdr(skb)->protocol == IPPROTO_IGMP) {
1974                         /* IGMPv1 (and broken IGMPv2 implementations sort of
1975                          * Cisco IOS <= 11.2(8)) do not put router alert
1976                          * option to IGMP packets destined to routable
1977                          * groups. It is very bad, because it means
1978                          * that we can forward NO IGMP messages.
1979                          */
1980                         struct sock *mroute_sk;
1981
1982                         mroute_sk = rcu_dereference(mrt->mroute_sk);
1983                         if (mroute_sk) {
1984                                 nf_reset(skb);
1985                                 raw_rcv(mroute_sk, skb);
1986                                 return 0;
1987                         }
1988                     }
1989         }
1990
1991         /* already under rcu_read_lock() */
1992         cache = ipmr_cache_find(mrt, ip_hdr(skb)->saddr, ip_hdr(skb)->daddr);
1993         if (cache == NULL) {
1994                 int vif = ipmr_find_vif(mrt, skb->dev);
1995
1996                 if (vif >= 0)
1997                         cache = ipmr_cache_find_any(mrt, ip_hdr(skb)->daddr,
1998                                                     vif);
1999         }
2000
2001         /*
2002          *      No usable cache entry
2003          */
2004         if (cache == NULL) {
2005                 int vif;
2006
2007                 if (local) {
2008                         struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
2009                         ip_local_deliver(skb);
2010                         if (skb2 == NULL)
2011                                 return -ENOBUFS;
2012                         skb = skb2;
2013                 }
2014
2015                 read_lock(&mrt_lock);
2016                 vif = ipmr_find_vif(mrt, skb->dev);
2017                 if (vif >= 0) {
2018                         int err2 = ipmr_cache_unresolved(mrt, vif, skb);
2019                         read_unlock(&mrt_lock);
2020
2021                         return err2;
2022                 }
2023                 read_unlock(&mrt_lock);
2024                 kfree_skb(skb);
2025                 return -ENODEV;
2026         }
2027
2028         read_lock(&mrt_lock);
2029         ip_mr_forward(net, mrt, skb, cache, local);
2030         read_unlock(&mrt_lock);
2031
2032         if (local)
2033                 return ip_local_deliver(skb);
2034
2035         return 0;
2036
2037 dont_forward:
2038         if (local)
2039                 return ip_local_deliver(skb);
2040         kfree_skb(skb);
2041         return 0;
2042 }
2043
2044 #ifdef CONFIG_IP_PIMSM
2045 /* called with rcu_read_lock() */
2046 static int __pim_rcv(struct mr_table *mrt, struct sk_buff *skb,
2047                      unsigned int pimlen)
2048 {
2049         struct net_device *reg_dev = NULL;
2050         struct iphdr *encap;
2051
2052         encap = (struct iphdr *)(skb_transport_header(skb) + pimlen);
2053         /*
2054          * Check that:
2055          * a. packet is really sent to a multicast group
2056          * b. packet is not a NULL-REGISTER
2057          * c. packet is not truncated
2058          */
2059         if (!ipv4_is_multicast(encap->daddr) ||
2060             encap->tot_len == 0 ||
2061             ntohs(encap->tot_len) + pimlen > skb->len)
2062                 return 1;
2063
2064         read_lock(&mrt_lock);
2065         if (mrt->mroute_reg_vif_num >= 0)
2066                 reg_dev = mrt->vif_table[mrt->mroute_reg_vif_num].dev;
2067         read_unlock(&mrt_lock);
2068
2069         if (reg_dev == NULL)
2070                 return 1;
2071
2072         skb->mac_header = skb->network_header;
2073         skb_pull(skb, (u8 *)encap - skb->data);
2074         skb_reset_network_header(skb);
2075         skb->protocol = htons(ETH_P_IP);
2076         skb->ip_summed = CHECKSUM_NONE;
2077         skb->pkt_type = PACKET_HOST;
2078
2079         skb_tunnel_rx(skb, reg_dev);
2080
2081         netif_rx(skb);
2082
2083         return NET_RX_SUCCESS;
2084 }
2085 #endif
2086
2087 #ifdef CONFIG_IP_PIMSM_V1
2088 /*
2089  * Handle IGMP messages of PIMv1
2090  */
2091
2092 int pim_rcv_v1(struct sk_buff *skb)
2093 {
2094         struct igmphdr *pim;
2095         struct net *net = dev_net(skb->dev);
2096         struct mr_table *mrt;
2097
2098         if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2099                 goto drop;
2100
2101         pim = igmp_hdr(skb);
2102
2103         mrt = ipmr_rt_fib_lookup(net, skb);
2104         if (IS_ERR(mrt))
2105                 goto drop;
2106         if (!mrt->mroute_do_pim ||
2107             pim->group != PIM_V1_VERSION || pim->code != PIM_V1_REGISTER)
2108                 goto drop;
2109
2110         if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2111 drop:
2112                 kfree_skb(skb);
2113         }
2114         return 0;
2115 }
2116 #endif
2117
2118 #ifdef CONFIG_IP_PIMSM_V2
2119 static int pim_rcv(struct sk_buff *skb)
2120 {
2121         struct pimreghdr *pim;
2122         struct net *net = dev_net(skb->dev);
2123         struct mr_table *mrt;
2124
2125         if (!pskb_may_pull(skb, sizeof(*pim) + sizeof(struct iphdr)))
2126                 goto drop;
2127
2128         pim = (struct pimreghdr *)skb_transport_header(skb);
2129         if (pim->type != ((PIM_VERSION << 4) | (PIM_REGISTER)) ||
2130             (pim->flags & PIM_NULL_REGISTER) ||
2131             (ip_compute_csum((void *)pim, sizeof(*pim)) != 0 &&
2132              csum_fold(skb_checksum(skb, 0, skb->len, 0))))
2133                 goto drop;
2134
2135         mrt = ipmr_rt_fib_lookup(net, skb);
2136         if (IS_ERR(mrt))
2137                 goto drop;
2138         if (__pim_rcv(mrt, skb, sizeof(*pim))) {
2139 drop:
2140                 kfree_skb(skb);
2141         }
2142         return 0;
2143 }
2144 #endif
2145
2146 static int __ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2147                               struct mfc_cache *c, struct rtmsg *rtm)
2148 {
2149         int ct;
2150         struct rtnexthop *nhp;
2151         struct nlattr *mp_attr;
2152         struct rta_mfc_stats mfcs;
2153
2154         /* If cache is unresolved, don't try to parse IIF and OIF */
2155         if (c->mfc_parent >= MAXVIFS)
2156                 return -ENOENT;
2157
2158         if (VIF_EXISTS(mrt, c->mfc_parent) &&
2159             nla_put_u32(skb, RTA_IIF, mrt->vif_table[c->mfc_parent].dev->ifindex) < 0)
2160                 return -EMSGSIZE;
2161
2162         if (!(mp_attr = nla_nest_start(skb, RTA_MULTIPATH)))
2163                 return -EMSGSIZE;
2164
2165         for (ct = c->mfc_un.res.minvif; ct < c->mfc_un.res.maxvif; ct++) {
2166                 if (VIF_EXISTS(mrt, ct) && c->mfc_un.res.ttls[ct] < 255) {
2167                         if (!(nhp = nla_reserve_nohdr(skb, sizeof(*nhp)))) {
2168                                 nla_nest_cancel(skb, mp_attr);
2169                                 return -EMSGSIZE;
2170                         }
2171
2172                         nhp->rtnh_flags = 0;
2173                         nhp->rtnh_hops = c->mfc_un.res.ttls[ct];
2174                         nhp->rtnh_ifindex = mrt->vif_table[ct].dev->ifindex;
2175                         nhp->rtnh_len = sizeof(*nhp);
2176                 }
2177         }
2178
2179         nla_nest_end(skb, mp_attr);
2180
2181         mfcs.mfcs_packets = c->mfc_un.res.pkt;
2182         mfcs.mfcs_bytes = c->mfc_un.res.bytes;
2183         mfcs.mfcs_wrong_if = c->mfc_un.res.wrong_if;
2184         if (nla_put(skb, RTA_MFC_STATS, sizeof(mfcs), &mfcs) < 0)
2185                 return -EMSGSIZE;
2186
2187         rtm->rtm_type = RTN_MULTICAST;
2188         return 1;
2189 }
2190
2191 int ipmr_get_route(struct net *net, struct sk_buff *skb,
2192                    __be32 saddr, __be32 daddr,
2193                    struct rtmsg *rtm, int nowait)
2194 {
2195         struct mfc_cache *cache;
2196         struct mr_table *mrt;
2197         int err;
2198
2199         mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2200         if (mrt == NULL)
2201                 return -ENOENT;
2202
2203         rcu_read_lock();
2204         cache = ipmr_cache_find(mrt, saddr, daddr);
2205         if (cache == NULL && skb->dev) {
2206                 int vif = ipmr_find_vif(mrt, skb->dev);
2207
2208                 if (vif >= 0)
2209                         cache = ipmr_cache_find_any(mrt, daddr, vif);
2210         }
2211         if (cache == NULL) {
2212                 struct sk_buff *skb2;
2213                 struct iphdr *iph;
2214                 struct net_device *dev;
2215                 int vif = -1;
2216
2217                 if (nowait) {
2218                         rcu_read_unlock();
2219                         return -EAGAIN;
2220                 }
2221
2222                 dev = skb->dev;
2223                 read_lock(&mrt_lock);
2224                 if (dev)
2225                         vif = ipmr_find_vif(mrt, dev);
2226                 if (vif < 0) {
2227                         read_unlock(&mrt_lock);
2228                         rcu_read_unlock();
2229                         return -ENODEV;
2230                 }
2231                 skb2 = skb_clone(skb, GFP_ATOMIC);
2232                 if (!skb2) {
2233                         read_unlock(&mrt_lock);
2234                         rcu_read_unlock();
2235                         return -ENOMEM;
2236                 }
2237
2238                 skb_push(skb2, sizeof(struct iphdr));
2239                 skb_reset_network_header(skb2);
2240                 iph = ip_hdr(skb2);
2241                 iph->ihl = sizeof(struct iphdr) >> 2;
2242                 iph->saddr = saddr;
2243                 iph->daddr = daddr;
2244                 iph->version = 0;
2245                 err = ipmr_cache_unresolved(mrt, vif, skb2);
2246                 read_unlock(&mrt_lock);
2247                 rcu_read_unlock();
2248                 return err;
2249         }
2250
2251         read_lock(&mrt_lock);
2252         if (!nowait && (rtm->rtm_flags & RTM_F_NOTIFY))
2253                 cache->mfc_flags |= MFC_NOTIFY;
2254         err = __ipmr_fill_mroute(mrt, skb, cache, rtm);
2255         read_unlock(&mrt_lock);
2256         rcu_read_unlock();
2257         return err;
2258 }
2259
2260 static int ipmr_fill_mroute(struct mr_table *mrt, struct sk_buff *skb,
2261                             u32 portid, u32 seq, struct mfc_cache *c, int cmd)
2262 {
2263         struct nlmsghdr *nlh;
2264         struct rtmsg *rtm;
2265         int err;
2266
2267         nlh = nlmsg_put(skb, portid, seq, cmd, sizeof(*rtm), NLM_F_MULTI);
2268         if (nlh == NULL)
2269                 return -EMSGSIZE;
2270
2271         rtm = nlmsg_data(nlh);
2272         rtm->rtm_family   = RTNL_FAMILY_IPMR;
2273         rtm->rtm_dst_len  = 32;
2274         rtm->rtm_src_len  = 32;
2275         rtm->rtm_tos      = 0;
2276         rtm->rtm_table    = mrt->id;
2277         if (nla_put_u32(skb, RTA_TABLE, mrt->id))
2278                 goto nla_put_failure;
2279         rtm->rtm_type     = RTN_MULTICAST;
2280         rtm->rtm_scope    = RT_SCOPE_UNIVERSE;
2281         if (c->mfc_flags & MFC_STATIC)
2282                 rtm->rtm_protocol = RTPROT_STATIC;
2283         else
2284                 rtm->rtm_protocol = RTPROT_MROUTED;
2285         rtm->rtm_flags    = 0;
2286
2287         if (nla_put_be32(skb, RTA_SRC, c->mfc_origin) ||
2288             nla_put_be32(skb, RTA_DST, c->mfc_mcastgrp))
2289                 goto nla_put_failure;
2290         err = __ipmr_fill_mroute(mrt, skb, c, rtm);
2291         /* do not break the dump if cache is unresolved */
2292         if (err < 0 && err != -ENOENT)
2293                 goto nla_put_failure;
2294
2295         return nlmsg_end(skb, nlh);
2296
2297 nla_put_failure:
2298         nlmsg_cancel(skb, nlh);
2299         return -EMSGSIZE;
2300 }
2301
2302 static size_t mroute_msgsize(bool unresolved, int maxvif)
2303 {
2304         size_t len =
2305                 NLMSG_ALIGN(sizeof(struct rtmsg))
2306                 + nla_total_size(4)     /* RTA_TABLE */
2307                 + nla_total_size(4)     /* RTA_SRC */
2308                 + nla_total_size(4)     /* RTA_DST */
2309                 ;
2310
2311         if (!unresolved)
2312                 len = len
2313                       + nla_total_size(4)       /* RTA_IIF */
2314                       + nla_total_size(0)       /* RTA_MULTIPATH */
2315                       + maxvif * NLA_ALIGN(sizeof(struct rtnexthop))
2316                                                 /* RTA_MFC_STATS */
2317                       + nla_total_size(sizeof(struct rta_mfc_stats))
2318                 ;
2319
2320         return len;
2321 }
2322
2323 static void mroute_netlink_event(struct mr_table *mrt, struct mfc_cache *mfc,
2324                                  int cmd)
2325 {
2326         struct net *net = read_pnet(&mrt->net);
2327         struct sk_buff *skb;
2328         int err = -ENOBUFS;
2329
2330         skb = nlmsg_new(mroute_msgsize(mfc->mfc_parent >= MAXVIFS, mrt->maxvif),
2331                         GFP_ATOMIC);
2332         if (skb == NULL)
2333                 goto errout;
2334
2335         err = ipmr_fill_mroute(mrt, skb, 0, 0, mfc, cmd);
2336         if (err < 0)
2337                 goto errout;
2338
2339         rtnl_notify(skb, net, 0, RTNLGRP_IPV4_MROUTE, NULL, GFP_ATOMIC);
2340         return;
2341
2342 errout:
2343         kfree_skb(skb);
2344         if (err < 0)
2345                 rtnl_set_sk_err(net, RTNLGRP_IPV4_MROUTE, err);
2346 }
2347
2348 static int ipmr_rtm_dumproute(struct sk_buff *skb, struct netlink_callback *cb)
2349 {
2350         struct net *net = sock_net(skb->sk);
2351         struct mr_table *mrt;
2352         struct mfc_cache *mfc;
2353         unsigned int t = 0, s_t;
2354         unsigned int h = 0, s_h;
2355         unsigned int e = 0, s_e;
2356
2357         s_t = cb->args[0];
2358         s_h = cb->args[1];
2359         s_e = cb->args[2];
2360
2361         rcu_read_lock();
2362         ipmr_for_each_table(mrt, net) {
2363                 if (t < s_t)
2364                         goto next_table;
2365                 if (t > s_t)
2366                         s_h = 0;
2367                 for (h = s_h; h < MFC_LINES; h++) {
2368                         list_for_each_entry_rcu(mfc, &mrt->mfc_cache_array[h], list) {
2369                                 if (e < s_e)
2370                                         goto next_entry;
2371                                 if (ipmr_fill_mroute(mrt, skb,
2372                                                      NETLINK_CB(cb->skb).portid,
2373                                                      cb->nlh->nlmsg_seq,
2374                                                      mfc, RTM_NEWROUTE) < 0)
2375                                         goto done;
2376 next_entry:
2377                                 e++;
2378                         }
2379                         e = s_e = 0;
2380                 }
2381                 spin_lock_bh(&mfc_unres_lock);
2382                 list_for_each_entry(mfc, &mrt->mfc_unres_queue, list) {
2383                         if (e < s_e)
2384                                 goto next_entry2;
2385                         if (ipmr_fill_mroute(mrt, skb,
2386                                              NETLINK_CB(cb->skb).portid,
2387                                              cb->nlh->nlmsg_seq,
2388                                              mfc, RTM_NEWROUTE) < 0) {
2389                                 spin_unlock_bh(&mfc_unres_lock);
2390                                 goto done;
2391                         }
2392 next_entry2:
2393                         e++;
2394                 }
2395                 spin_unlock_bh(&mfc_unres_lock);
2396                 e = s_e = 0;
2397                 s_h = 0;
2398 next_table:
2399                 t++;
2400         }
2401 done:
2402         rcu_read_unlock();
2403
2404         cb->args[2] = e;
2405         cb->args[1] = h;
2406         cb->args[0] = t;
2407
2408         return skb->len;
2409 }
2410
2411 #ifdef CONFIG_PROC_FS
2412 /*
2413  *      The /proc interfaces to multicast routing :
2414  *      /proc/net/ip_mr_cache & /proc/net/ip_mr_vif
2415  */
2416 struct ipmr_vif_iter {
2417         struct seq_net_private p;
2418         struct mr_table *mrt;
2419         int ct;
2420 };
2421
2422 static struct vif_device *ipmr_vif_seq_idx(struct net *net,
2423                                            struct ipmr_vif_iter *iter,
2424                                            loff_t pos)
2425 {
2426         struct mr_table *mrt = iter->mrt;
2427
2428         for (iter->ct = 0; iter->ct < mrt->maxvif; ++iter->ct) {
2429                 if (!VIF_EXISTS(mrt, iter->ct))
2430                         continue;
2431                 if (pos-- == 0)
2432                         return &mrt->vif_table[iter->ct];
2433         }
2434         return NULL;
2435 }
2436
2437 static void *ipmr_vif_seq_start(struct seq_file *seq, loff_t *pos)
2438         __acquires(mrt_lock)
2439 {
2440         struct ipmr_vif_iter *iter = seq->private;
2441         struct net *net = seq_file_net(seq);
2442         struct mr_table *mrt;
2443
2444         mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2445         if (mrt == NULL)
2446                 return ERR_PTR(-ENOENT);
2447
2448         iter->mrt = mrt;
2449
2450         read_lock(&mrt_lock);
2451         return *pos ? ipmr_vif_seq_idx(net, seq->private, *pos - 1)
2452                 : SEQ_START_TOKEN;
2453 }
2454
2455 static void *ipmr_vif_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2456 {
2457         struct ipmr_vif_iter *iter = seq->private;
2458         struct net *net = seq_file_net(seq);
2459         struct mr_table *mrt = iter->mrt;
2460
2461         ++*pos;
2462         if (v == SEQ_START_TOKEN)
2463                 return ipmr_vif_seq_idx(net, iter, 0);
2464
2465         while (++iter->ct < mrt->maxvif) {
2466                 if (!VIF_EXISTS(mrt, iter->ct))
2467                         continue;
2468                 return &mrt->vif_table[iter->ct];
2469         }
2470         return NULL;
2471 }
2472
2473 static void ipmr_vif_seq_stop(struct seq_file *seq, void *v)
2474         __releases(mrt_lock)
2475 {
2476         read_unlock(&mrt_lock);
2477 }
2478
2479 static int ipmr_vif_seq_show(struct seq_file *seq, void *v)
2480 {
2481         struct ipmr_vif_iter *iter = seq->private;
2482         struct mr_table *mrt = iter->mrt;
2483
2484         if (v == SEQ_START_TOKEN) {
2485                 seq_puts(seq,
2486                          "Interface      BytesIn  PktsIn  BytesOut PktsOut Flags Local    Remote\n");
2487         } else {
2488                 const struct vif_device *vif = v;
2489                 const char *name =  vif->dev ? vif->dev->name : "none";
2490
2491                 seq_printf(seq,
2492                            "%2Zd %-10s %8ld %7ld  %8ld %7ld %05X %08X %08X\n",
2493                            vif - mrt->vif_table,
2494                            name, vif->bytes_in, vif->pkt_in,
2495                            vif->bytes_out, vif->pkt_out,
2496                            vif->flags, vif->local, vif->remote);
2497         }
2498         return 0;
2499 }
2500
2501 static const struct seq_operations ipmr_vif_seq_ops = {
2502         .start = ipmr_vif_seq_start,
2503         .next  = ipmr_vif_seq_next,
2504         .stop  = ipmr_vif_seq_stop,
2505         .show  = ipmr_vif_seq_show,
2506 };
2507
2508 static int ipmr_vif_open(struct inode *inode, struct file *file)
2509 {
2510         return seq_open_net(inode, file, &ipmr_vif_seq_ops,
2511                             sizeof(struct ipmr_vif_iter));
2512 }
2513
2514 static const struct file_operations ipmr_vif_fops = {
2515         .owner   = THIS_MODULE,
2516         .open    = ipmr_vif_open,
2517         .read    = seq_read,
2518         .llseek  = seq_lseek,
2519         .release = seq_release_net,
2520 };
2521
2522 struct ipmr_mfc_iter {
2523         struct seq_net_private p;
2524         struct mr_table *mrt;
2525         struct list_head *cache;
2526         int ct;
2527 };
2528
2529
2530 static struct mfc_cache *ipmr_mfc_seq_idx(struct net *net,
2531                                           struct ipmr_mfc_iter *it, loff_t pos)
2532 {
2533         struct mr_table *mrt = it->mrt;
2534         struct mfc_cache *mfc;
2535
2536         rcu_read_lock();
2537         for (it->ct = 0; it->ct < MFC_LINES; it->ct++) {
2538                 it->cache = &mrt->mfc_cache_array[it->ct];
2539                 list_for_each_entry_rcu(mfc, it->cache, list)
2540                         if (pos-- == 0)
2541                                 return mfc;
2542         }
2543         rcu_read_unlock();
2544
2545         spin_lock_bh(&mfc_unres_lock);
2546         it->cache = &mrt->mfc_unres_queue;
2547         list_for_each_entry(mfc, it->cache, list)
2548                 if (pos-- == 0)
2549                         return mfc;
2550         spin_unlock_bh(&mfc_unres_lock);
2551
2552         it->cache = NULL;
2553         return NULL;
2554 }
2555
2556
2557 static void *ipmr_mfc_seq_start(struct seq_file *seq, loff_t *pos)
2558 {
2559         struct ipmr_mfc_iter *it = seq->private;
2560         struct net *net = seq_file_net(seq);
2561         struct mr_table *mrt;
2562
2563         mrt = ipmr_get_table(net, RT_TABLE_DEFAULT);
2564         if (mrt == NULL)
2565                 return ERR_PTR(-ENOENT);
2566
2567         it->mrt = mrt;
2568         it->cache = NULL;
2569         it->ct = 0;
2570         return *pos ? ipmr_mfc_seq_idx(net, seq->private, *pos - 1)
2571                 : SEQ_START_TOKEN;
2572 }
2573
2574 static void *ipmr_mfc_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2575 {
2576         struct mfc_cache *mfc = v;
2577         struct ipmr_mfc_iter *it = seq->private;
2578         struct net *net = seq_file_net(seq);
2579         struct mr_table *mrt = it->mrt;
2580
2581         ++*pos;
2582
2583         if (v == SEQ_START_TOKEN)
2584                 return ipmr_mfc_seq_idx(net, seq->private, 0);
2585
2586         if (mfc->list.next != it->cache)
2587                 return list_entry(mfc->list.next, struct mfc_cache, list);
2588
2589         if (it->cache == &mrt->mfc_unres_queue)
2590                 goto end_of_list;
2591
2592         BUG_ON(it->cache != &mrt->mfc_cache_array[it->ct]);
2593
2594         while (++it->ct < MFC_LINES) {
2595                 it->cache = &mrt->mfc_cache_array[it->ct];
2596                 if (list_empty(it->cache))
2597                         continue;
2598                 return list_first_entry(it->cache, struct mfc_cache, list);
2599         }
2600
2601         /* exhausted cache_array, show unresolved */
2602         rcu_read_unlock();
2603         it->cache = &mrt->mfc_unres_queue;
2604         it->ct = 0;
2605
2606         spin_lock_bh(&mfc_unres_lock);
2607         if (!list_empty(it->cache))
2608                 return list_first_entry(it->cache, struct mfc_cache, list);
2609
2610 end_of_list:
2611         spin_unlock_bh(&mfc_unres_lock);
2612         it->cache = NULL;
2613
2614         return NULL;
2615 }
2616
2617 static void ipmr_mfc_seq_stop(struct seq_file *seq, void *v)
2618 {
2619         struct ipmr_mfc_iter *it = seq->private;
2620         struct mr_table *mrt = it->mrt;
2621
2622         if (it->cache == &mrt->mfc_unres_queue)
2623                 spin_unlock_bh(&mfc_unres_lock);
2624         else if (it->cache == &mrt->mfc_cache_array[it->ct])
2625                 rcu_read_unlock();
2626 }
2627
2628 static int ipmr_mfc_seq_show(struct seq_file *seq, void *v)
2629 {
2630         int n;
2631
2632         if (v == SEQ_START_TOKEN) {
2633                 seq_puts(seq,
2634                  "Group    Origin   Iif     Pkts    Bytes    Wrong Oifs\n");
2635         } else {
2636                 const struct mfc_cache *mfc = v;
2637                 const struct ipmr_mfc_iter *it = seq->private;
2638                 const struct mr_table *mrt = it->mrt;
2639
2640                 seq_printf(seq, "%08X %08X %-3hd",
2641                            (__force u32) mfc->mfc_mcastgrp,
2642                            (__force u32) mfc->mfc_origin,
2643                            mfc->mfc_parent);
2644
2645                 if (it->cache != &mrt->mfc_unres_queue) {
2646                         seq_printf(seq, " %8lu %8lu %8lu",
2647                                    mfc->mfc_un.res.pkt,
2648                                    mfc->mfc_un.res.bytes,
2649                                    mfc->mfc_un.res.wrong_if);
2650                         for (n = mfc->mfc_un.res.minvif;
2651                              n < mfc->mfc_un.res.maxvif; n++) {
2652                                 if (VIF_EXISTS(mrt, n) &&
2653                                     mfc->mfc_un.res.ttls[n] < 255)
2654                                         seq_printf(seq,
2655                                            " %2d:%-3d",
2656                                            n, mfc->mfc_un.res.ttls[n]);
2657                         }
2658                 } else {
2659                         /* unresolved mfc_caches don't contain
2660                          * pkt, bytes and wrong_if values
2661                          */
2662                         seq_printf(seq, " %8lu %8lu %8lu", 0ul, 0ul, 0ul);
2663                 }
2664                 seq_putc(seq, '\n');
2665         }
2666         return 0;
2667 }
2668
2669 static const struct seq_operations ipmr_mfc_seq_ops = {
2670         .start = ipmr_mfc_seq_start,
2671         .next  = ipmr_mfc_seq_next,
2672         .stop  = ipmr_mfc_seq_stop,
2673         .show  = ipmr_mfc_seq_show,
2674 };
2675
2676 static int ipmr_mfc_open(struct inode *inode, struct file *file)
2677 {
2678         return seq_open_net(inode, file, &ipmr_mfc_seq_ops,
2679                             sizeof(struct ipmr_mfc_iter));
2680 }
2681
2682 static const struct file_operations ipmr_mfc_fops = {
2683         .owner   = THIS_MODULE,
2684         .open    = ipmr_mfc_open,
2685         .read    = seq_read,
2686         .llseek  = seq_lseek,
2687         .release = seq_release_net,
2688 };
2689 #endif
2690
2691 #ifdef CONFIG_IP_PIMSM_V2
2692 static const struct net_protocol pim_protocol = {
2693         .handler        =       pim_rcv,
2694         .netns_ok       =       1,
2695 };
2696 #endif
2697
2698
2699 /*
2700  *      Setup for IP multicast routing
2701  */
2702 static int __net_init ipmr_net_init(struct net *net)
2703 {
2704         int err;
2705
2706         err = ipmr_rules_init(net);
2707         if (err < 0)
2708                 goto fail;
2709
2710 #ifdef CONFIG_PROC_FS
2711         err = -ENOMEM;
2712         if (!proc_create("ip_mr_vif", 0, net->proc_net, &ipmr_vif_fops))
2713                 goto proc_vif_fail;
2714         if (!proc_create("ip_mr_cache", 0, net->proc_net, &ipmr_mfc_fops))
2715                 goto proc_cache_fail;
2716 #endif
2717         return 0;
2718
2719 #ifdef CONFIG_PROC_FS
2720 proc_cache_fail:
2721         remove_proc_entry("ip_mr_vif", net->proc_net);
2722 proc_vif_fail:
2723         ipmr_rules_exit(net);
2724 #endif
2725 fail:
2726         return err;
2727 }
2728
2729 static void __net_exit ipmr_net_exit(struct net *net)
2730 {
2731 #ifdef CONFIG_PROC_FS
2732         remove_proc_entry("ip_mr_cache", net->proc_net);
2733         remove_proc_entry("ip_mr_vif", net->proc_net);
2734 #endif
2735         ipmr_rules_exit(net);
2736 }
2737
2738 static struct pernet_operations ipmr_net_ops = {
2739         .init = ipmr_net_init,
2740         .exit = ipmr_net_exit,
2741 };
2742
2743 int __init ip_mr_init(void)
2744 {
2745         int err;
2746
2747         mrt_cachep = kmem_cache_create("ip_mrt_cache",
2748                                        sizeof(struct mfc_cache),
2749                                        0, SLAB_HWCACHE_ALIGN | SLAB_PANIC,
2750                                        NULL);
2751         if (!mrt_cachep)
2752                 return -ENOMEM;
2753
2754         err = register_pernet_subsys(&ipmr_net_ops);
2755         if (err)
2756                 goto reg_pernet_fail;
2757
2758         err = register_netdevice_notifier(&ip_mr_notifier);
2759         if (err)
2760                 goto reg_notif_fail;
2761 #ifdef CONFIG_IP_PIMSM_V2
2762         if (inet_add_protocol(&pim_protocol, IPPROTO_PIM) < 0) {
2763                 pr_err("%s: can't add PIM protocol\n", __func__);
2764                 err = -EAGAIN;
2765                 goto add_proto_fail;
2766         }
2767 #endif
2768         rtnl_register(RTNL_FAMILY_IPMR, RTM_GETROUTE,
2769                       NULL, ipmr_rtm_dumproute, NULL);
2770         return 0;
2771
2772 #ifdef CONFIG_IP_PIMSM_V2
2773 add_proto_fail:
2774         unregister_netdevice_notifier(&ip_mr_notifier);
2775 #endif
2776 reg_notif_fail:
2777         unregister_pernet_subsys(&ipmr_net_ops);
2778 reg_pernet_fail:
2779         kmem_cache_destroy(mrt_cachep);
2780         return err;
2781 }